Release 1.0 - December, 2005

R l l 0
ESTHER DYSON'S QUARTERLY REPORT e C a S C o

VOLUME 24, NO. 2 | DECEMBER 2005 | www.releasel-0.com

When 2.0: Time and Timing

INSIDE

WHEN 2.0: TIME & TIMING 1

Adding Value to Time 2
Why now? A short history

My Schedule, Your Schedule 6
Box: Three layers of scheduled time
OSAF’s Chandler: The schedule knows
IBM: Unified Activity Management
Trumba: Shared lives
Airena’s AirSet: Precision planning
Zimbra: Total control by owners

Scheduling: The Transaction Side 17
Renkoo: My time or yours?

TimeBridge: Game theory for meetings
Microsoft: Time of change
Negotiating changes: Bob and Carol and...

Events on Parade 25
Upcoming.org: Up and coming
EVDB: Eventfully yours
Zvents: Suddenly last summer
Columba Systems: Time and money
Seraja: A million single points in time
Demand ID: Demand it and he will come
Meetup.com: Bowling together
Technorati: Net future value

Adding Value With Time 42
Worktopia: Office arbitrage
CDC: Time is money life
Sana Security: Time and security
Riya: Eyes only for you

Time and Intentions 52
Revenue Science: Take me to your goldmine
Dan Doman: Intentions and inclinations

The End of the Beginning 58

Resources & Contact Information 59

Keep an eye out for your invitation
to PC Forum 2006, March 12 to 14,
at La Costa Resort and Spa.

BY ESTHER DYSON

Time is all we’ve got. Our challenge is allocating that time, intersecting
our time with that of others, managing the disposition over time of the
resources we control. Now at last we’re getting better tools to help us.

Time itself is abstract, but it takes on value as a measure of unique,
un-tradable things: Juan’s presence, the use of Alice’s spare apartment,
the time of a particular doctor or the attention of a specific audience.
One night is nothing; one night with you, per Elvis Presley, would
make my dreams come true.

But computers know nothing of this, even though time is intrinsic to
their operation and they can measure it with precision. Its meaning is
as foreign to them as the meaning of a mile or a kilometer. They don’t
understand how people value time, nor how time changes value — both
its own value, and the value of the things it measures.

Over time (to coin a phrase), many applications have incorporated
time — everything from length of employment to depreciation curves
and net present value calculations. But developers are just beginning
to understand the meaning of personal time. Most obvious is calen-
dars, scheduling, events, resource allocation over time (aka project
management). But there are also less obvious ways time matters in
software: how people work and think over time; how human relation-
ships, article relevance, and purchase intentions and other commercial
considerations change over time; how time patterns infuse a variety of
applications; and how a sense of timing can improve the utility of
everything from search results to social-network-driven tools.

And finally, there’s the future. We can predict it, or we can start pay-
ing intelligent attention to users’ stated intentions and desires. We

{ continued on page 2 }

Release 1.0® (ISSN 1047-935X) is
published quarterly by CNET Networks,
104 Fifth Avenue, New York, NY 10011-
6987; 1 (212) 924-8800; fax, 1 (212)
924-0240; www.release1-O.com. It cov-
ers the worlds of information technolo-
gy and the Internet, including wireless
communications, security, business
models, online services, tracking sys-
tems, identity management and other
unpredictable topics. ..and the policy
issues they raise.

EDITOR: Esther Dyson
(edyson@release1-0.com)

PUBLISHER: Daphne Kis
(daphne@release1-0.com)

MANAGING EDITOR: Christina Koukkos
(christina@release1-O0.com)

CONTRIBUTING WRITERS: Dan Farber
(dan.farber@cnet.com), Dan
Gillmor (dan@gillmor.com),
Steven Johnson (stevenberlin-
johnson@earthlink.net),Scott
Kirsner (kirsner@pobox.com),
Clay Shirky (clay@shirky.com),
Dave Weinberger
(self@evident.com)

CIRCULATION MANAGER: Brodie Crawford
(brodie@release1-0.com)

SYSTEMS MANAGER: Geoff Clarke
(geoff@release1-0.com)

Copyright © 2005, CNET Networks,
Inc. All rights reserved. No material in
this publication may be reproduced
without written permission; however,
we gladly arrange for reprints, bulk
orders or site licenses. Subscriptions
cost $285 per year.

2 RELEASE 1.0

can detect threats or opportunities or simply trends that aren’t yet
visible and thereby improve our timing — the better to generate a
response, avert a threat or catch a customer.

In this issue, we first consider traditional time-based applications
and tools, starting with calendars and including scheduling tools
and events databases. We show that while time can measure, a better
way to manage may be by first defining discrete activities that fill
our time. In the second part, we consider a number of applications
that use time in order to perform other functions better — ranging
from face recognition to behavioral targeting.

One message becomes clear: The online world needs to get better at
time-stamping content and activities and at standards for represent-
ing time and events — both times and durations, and all the patterns
in time: speed, decay, growth, recurrence, (changing) frequency of
events. Many applications spend lots of effort on interoperability
that wouldn't be necessary if we had standards.

Adding Value to Time: Calendars and
Schedules - and Maps

You can’t create time. You can only steal it, reallocate it, use it or waste
it. Every living and non-living thing has its span of existence. Calendars
are a way to map the path of each thing through time. Scheduling tools
try to create intersections across those paths. And events databases show
possible events through which you could pass.

There are lots of parallels between calendars and maps. Both enable
us to visualize — maps in two dimensions, and calendars in one. As
with maps, the visualization of a single aspect of a geography ora
single person’s schedule is easy; the tough part is when you combine
two or more layers in a map — for example, a street map plus satellite
images plus, say, the layout of an underlying sewer system. Geo-
coding serves to tie the different layers together. In theory, it’s all
clean and tidy, but the reality is much messier.

WWW.RELEASE1-0.COM

The same is true of schedules: No man (‘s calendar) is an island; the default is figur-
ing out how to share events. The challenges of mingling schedules are fairly easy
when it concerns your typical corporate environment, with a (potentially large)
number of users all running off a single Exchange server. But now, more people are
trying to schedule events outside the confines of that controlled environment.

Any single person’s life should be quite simple, but in fact, lots of related things hap-
pen in parallel. Most things that take time involve other people; otherwise, they
wouldn’t be so hard to schedule! And your own schedule may contain commit-
ments, possibilities, meetings requested by others, deadlines and so on, which
together represent a multiplicity of possible, contradictory futures. (Divining possi-
ble pasts is a task called forensics.)

Indeed, the issues around scheduling also have to do with social algorithms: Whose
busy-ness takes precedence over who else’s? Can Juan see Alice’s calendar, or only
that she is busy? What if she’s busy for Juan but not for, say, her boss or a valued
client? The current set of scheduling tools is adding social affordances for people
who value their time. What should be the defaults for disclosure? And how should
they change over time?

There is also secret time, which may show on a calendar as revealing gaps — akin to
contacts with whom you have no visible (or anyway, saved) communication. That’s
something we talked about with Microsoft CTO Ray Ozzie, who works in the same
building as Bill Gates; that building is the only one whose layout is not freely avail-
able over the Microsoft intranet.

Finally, there are subtle cues to take into account. Schedule a meeting at 3 pm, and
people will assume they’ve got an hour. Schedule it at 4.40, and they'll take 20 min-
utes. But schedule it at 4.50 and they'll be confused — until you explain that you're
serious and that you can actually handle their “issues” in 5 minutes.

Something to watch over me

The best applications would watch what the user does and how she allocates her
time, and then start predicting those decisions. Suddenly, instead of making deci-
sions, the user would need only to correct the wrong ones.

Calendar applications would also do well to take into account all the high-level

objects associated with each calendar item: participants, resources (a conference
room), documents, e-mails, associated expenses, tickets or reservations. How can

DECEMBER 2005

RELEASE 1.0

RELEASE 1.0

these be represented and linked to or from? If Juan meets with Alice and puts her in
his calendar each time, shouldn’t Alice be a defined object in Juan’s calendar. . .so
that later he can look her up in his address book and find references to each time he
met with her? (That could be very helpful to a prosecutor when Juan gets into trou-
ble, but also useful to Juan himself or to a corporate application trying to find out
the strength of his connection to Alice — whether in order to get an introduction to
Alice, or to find out who in the company knows about the top-secret Doodle project
once Juan leaves.)

This is what IBM calls activity management. It’s in its infancy, but in the end, we
can’t manage time, we can manage only what we do with our time. Representing
activities in a meaningful, goal-focused way that allows us to coordinate all the
resources we need for any particular task will be a huge win.

Finally, most schedulers don’t handle location (other than as a descriptor) — say, that
the series of meetings in San Francisco should be contiguous, and that the meetings
in Palo Alto should also happen contiguously. That’s a task on their to-do list.

Why now? A short history

Modern calendar times began with the launch of Microsoft Outlook, with Office 97.
Now in its fifth release (as Office Outlook 2003), it has about 300 million users, most
of them tethered to a corporate Exchange server. It’s worth noting that Outlook has
a profusion of functions, but most people don’t know about them. Ease of use and
intuitiveness will be key to making the next generation of time- and activity-man-
agement tools broadly used — and useful.

In addition, Lotus Notes has about 120 million users. (SEE RELEASE 1.0, JUNE 2004).
However, most individuals use electronic calendars for business schedules but not for
personal matters — odd considering how “personal” a tool a calendar is. In particular,
says Joyce Park, who researched the market before co-founding Renkoo (Pace 18), few
women use electronic calendars, even though they tend to be responsible for families’
shared calendars. There just hasn’t been a compelling offering. For example, we per-
sonally use Microsoft Word to keep our calendar; were waiting for something good
enough that we will never have to switch again. There was a brief flurry of more con-
sumer-oriented calendar offerings along with a variety of other applications during
the dot-com boom, but few have survived. Most are now mostly neglected offerings
on big portals. When.com was absorbed by AOL; Jump.com lives on as MSN

WWW.RELEASE1-0.COM

WHEN 2.0 WORKSHOP

The companies and individuals profiled here will be partic- beginning of a long-term conversation that will have a vis-
ipating in Release 1.0's “When 2.0: Time and timing” work- ible, public manifestation on December 6, 2005.
shop at Stanford University. This newsletter is just the For more information visit www.release1-O.com/events.

Calendar. Part of the problem — now being addressed — is that none of them play well
together — and in the end most scheduling requires coordination with other people.

On the scheduling side, Interactive Corp. owns Evite, which has tens of millions of
users, while Yahoo! Groups includes a group calendar. ..which doesn’t communicate
especially well with Yahoo!’s own MyCalendar. Online scheduling application Time-
Dance got to 1 million users a month in 2001, and then fizzled as its new manage-
ment tried to generate revenues by selling greeting cards.

But now the field is heating up again.

On the private side, more people are interacting with more other people online, try-
ing to reconcile busy schedules and to set up meetings. Now they are looking for a
repository for all the as-yet unstructured data they are creating and all the scheduling
e-mails they are sending back and forth. They want to share schedule information
with others, but selectively, so that they can protect both their time and their privacy.

On the public side, new event portals are aggregating event announcements from
the Web and offering RSS feeds. Users need something to import those feeds into
that will know what to do with them — in short, a calendar.

Web publishers and calendar applications are beginning to use microformats and
standards to represent events. But so far, the information in most calendars is strings
of text, rather than explicit objects — contacts, locations, venues, resources, clients,
business expenses or time sheets — that can be tied to projects, goals and the like.
Books have ISBN numbers, planes have tail numbers. . .but there’s no registry of
venues, let alone of events themselves.

And at a completely different level of abstraction, IBM and the Open Source Ap-
plications Foundation (OSAF) are beginning to define ways to represent and man-
age “activities” — not times and dates, but collections of events, content and people
that are linked by a common endeavor, whether it be the building of a cathedral, the
election of a dark-horse candidate or the generation of next month’s sales forecast.
Time is a measure and an organizing principle; activity is, well, the purpose of it all.
But there’s only one stream of time that you can represent simply in a calendar or

DECEMBER 2005 RELEASE 1.0 5

RELEASE 1.0

timeline (such as the timeline in Outlook’s Journal view, which has some notion of
collecting data into activities), whereas almost everyone has multiple activities.

Yes, that threatens greater complexity, but it’s just a reflection of reality. Tools that
actually simplify the complexity and that automate routine work — rather than just
share data — may provide the payoff that will get people to use these tools.

The market emerges

Different start-ups are focusing on different facets of these puzzles, including inter-
operability of their tools across platforms such as cell phones and PDAs. The union
of scheduling and mail — still the primary communication tool for setting up
appointments — is a development we discussed at length in our June 2004 issue on
Meta-mail. At the time, most of the discussion was theoretical, with pointers to bits
of functionality here and there, plus IBM/Lotus’s more-inclusive but closed-group
Unified Activity Management (see pace 9). Now things are beginning to happen.

My Schedule, Your Schedule

Four leading independent calendar contenders currently are Chandler (being devel-
oped in public, together with the Cosmo server, as a project of OSAF), Trumba
(which debuted at last March’s PC Forum; SEE RELEASE 1.0, MARCH 2005), Airena’s
AirSet and Zimbra.

Chandler is a client and can share data with other clients through its own server,
Cosmo, whereas Trumba’s OneCalendar is a Web service that works with both
browser-based calendar applications (Yahoo! Calendar, MSN Calendar) and desktop
calendars (Outlook, Apple iCal) and is more focused on providing schedule-pub-
lishing and -sharing tools for organizations and group organizers. AirSet offers a
richer but more complex user experience and back-end than Trumba — it supports
overall group activities (contacts as well as scheduling) — while Zimbra is a mail-cen-
tered server that supports Outlook and other clients (replacing Exchange); it can
recognize events and contacts (among other things) in mail, and can semi-automati-
cally insert them into the user’s Outlook calendar.

Finally, there is the management of activities as well as time. In Chandler, it’s just a

new capability, offered in a lightweight way; in IBM’s Unified Activity Management
(UAM) project, it’s the center of attention.

WWW.RELEASE1-0.COM

THE THREE LAYERS OF SCHEDULED TIME

There are three intertwined layers of schedules:

Private/personal, which may be shared as a
whole or in part with other private calendars. Related
tools enable individuals to negotiate meetings, with vary-
ing levels of disclosure of their own “agendas.”

Group, where an organizer publishes a schedule
to a group, whose members may or may not be able to add
events themselves. Group organizers may send invitations
or - at their own risk - they may try to get the group to
come up with a consensus on time and place.

Public, which contains events/schedules pub-
lished for download by unaffiliated users. It's up to individ-

OSAF's Chandler: The schedule knows

uals to fit these events into their own lives (though they
may coordinate as groups trying to decide, for example,
which showing of “Good Night, and Good Luck" to see.

And then of course there's the latent layer of all
events mentioned on the Web, with no formatting or struc-
ture, from brief blog references to a party at someone's
house to hundreds of text announcements on websites
designed for human rather than machine consumption. In
the middle of this there are messages about events, rang-
ing from formally structured Evites or Outlook appoint-
ment requests to casual mails or IMs about “lunch
Thursday in the restaurant near Alice's office?"

Chandler, in the works at OSAF for four years, became the cornerstone of a suite of

personal tools. Says lead designer Lisa Dusseault, “We saw what happened to our sis-

ter non-profit, Level Playing Field Institute. It was a small organization falling into a

black hole of calendaring — unable to hire a person just to manage an Exchange serv-

er and finding other calendar software unusable.”

Chandler is a desktop application that can share data with other applications using

Cosmo, a general information-sharing server that handles other content as well as
schedule data. There’s also a Web-based user tool called Scooby that offers similar

functionality directly in and out of Cosmo. They all use the same data represented in
CalDAV, the IETF-developed calendar language/protocol. CalDAV supports full
sharing with multiple readers and writers and fine-grained access control. It also

scales, so that arbitrarily large numbers of events (for instance, a full BART schedule

or an entire year’s television programming) can be efficiently shared. CalDAV also

supports enterprise use cases including the ability to view somebody’s busy time and

browse through calendars. Most importantly, all this works across the Internet, with

native HTTP URLs for every calendar and each event.

Dusseault led the charge for CalDAV’s development as an individual even before she
joined OSAF in March 2004. CalDAV potentially brings capabilities equivalent to
those of Exchange, Notes or some of IBM’s office tools to a much broader user base.

It is already supported by Mozilla Lightning, Oracle and Novell.

And it talks to — or at least offers the possibility to talk to — cell phones and PDAs.
Notes Dusseault: “It doesn’t make sense to synchronize your calendar from your PC

to your PDA. The time when you most want up-to-date schedule information on

your PDA is precisely when youre far away from your laptop and you can’t synch.

DECEMBER 2005

RELEASE 1.0 7

RELEASE 1.0

}
a
m)
xi

andler

Fie Edit View Item Colection Test Help

y y N
x åa zZ > À z
All Mail Tasks |Calendar|! SyncAll New Send
My calendar
-gr <> November 2005 [sesen [vy] @ v(m) a a
Week | Sun20| Mon... | Tue 22| We... 3 |< | Afternoon tour, Library of Congress
| Washington D.C.
[ all-day
3 Europe/London m
breakfast on Capitol Hill Europe/Paris _ starts | 11/23/05 at | 1:00 PM
l a Asia/Shanghai
Asia/Calcutta ends | 11/23/05 at | 4:30 PM
Australia/Sydne
11 time zone [US/Eastern x]
12 status |Confirmed x]
occurs |Once
fee (I) aam [none f]
Afternoon tour, Library of arm (None
4 Today » 2f Congress Se
a
November 2005 E~
SMTWTFS 3
12345
6 7 8 9 10 1112
13 14 15.16 17 18 19 ’ =
20 21 22123 24 25 26
27 28 29 30 5
December 2005 6
SMTWTFS
i 2 3 z
4 5678910 7:00 PM Ls
11 12 13 14 15 16 17 Flight home - Dulles to SFO
18 19 20 21 22 23 24 sij
25 26 27 28 29 30 31 v

—|¥ | Appears in: My items

For us, aside from the basic architecture supporting activity management and openness, the most satisfying spe-

cific feature of Chandler (and in AirSet and a number of other mash-ups) is that it handles time zones just the
way you'd expect. Your calendar reflects the default time zone you specify; the individual events in it reflect the
time zone they are in. So for example, if you're in Washington, DC, your meetings will default to EST (if that’s
how you have set things up). But if you add a 1.15 pm PST call, it will read 1.15 pm PST and will show up
placed at 4.15 pm on the calendar grid. When you get to California and shift to PST, your past appointments
will show up placed three hours off, now that you are back in PST.

The only other thing we would add is a color wash, so that you could specify different time zones across part of a
multi-day calendar during which you are traveling and have them show up in a different color. That is, the time
you spend in EST could be white and the time in PST a light blue.

Modern PDAs with connectivity should be able to access the calendar server directly
and selectively — getting current information only.”

The second big innovation in Chandler is its ability to link different kinds of data
around the user’s activities — or what it calls “collections.” For example, says
Dusseault, any incoming e-mail — not just specially formatted invitations as in
Outlook — can be turned into an event. However, the user must make that link, while
a formatted Outlook invitation already “knows” what it is. So, says Dusseault, “If
Jane sends me e-mail saying TIl try to call you Sunday, I can turn that into an event
on my calendar for Sunday. When I look at my day or weekend plans, I'll see that as a

WWW.RELEASE1-0.COM

time-related piece of information. Note that this turns the e-mail into a calendar

event but it remains an e-mail too — it’s still also the message from Jane.”

Events, e-mails and tasks can live together in the same collection — or activity.

Activities are represented not just within a single user’s Chandler
schedule, but alsocan be replicated in Cosmo, so they can be shared
across users. By default, all members of an activity can see all of the
related items.

In addition to collections, Chandler/Cosmo also support tags — what
an earlier generation of mail users might have called “labels,” and a
more flexible option than hierarchical folders. Any data item — not
just a message — can have one or more tags. This enables powerful
search across the combined applications.

OSAF is announcing an experimentally usable, but early beta
version of Chandler (“download this only if youre willing to help us
fix the bugs”) early this month (December). As for Cosmo, says
Dusseault, “Were going through the glitches of hosting it ourselves
first.” Given that it’s free, Chandler/Cosmo has good hopes of
attracting individual users and groups of those users. “We don’t

OSAF INFO

Headquarters: San Francisco, CA

Founded: Summer 2001

Employees: 24

Funding: $7.75 million from Mitchell
Kapor, the Andrew W. Mellon
Foundation and Common Solutions
Group, plus small donations from
private individuals

Key metric: Chandler 0.6 due out in
December 2005 will be “experimen-
tally usable” for individual and

group calendaring

URL: chandler.osafoundation.org

need to fund a marketing organization,” says Kapor. “All we need is about $3 million

a year for the development team.” That could come from the current funders,

individuals or new sources.

IBM Unified Activity Management: Managing activities, not time

As we described last June (SEE RELEASE 1.0, JUNE 2004), IBM Research is doing

groundbreaking work in what can best be described as personal workflow — or activ-
ity management. Says Tom Moran, an IBM Distinguished Engineer and leader of the
Unified Activity Management project: “I usually contrast flexible ‘activity’ adapted to
local circumstances with rigidly determined ‘workflow, ’ and I prefer ‘interpersonal’
to ‘personal. Lotus has made clear to us that personal productivity tools per se are of
no value to their enterprise customers, and that their concern is with ‘organizational
productivity’ We take that to mean the informal collaborative work that business
people actually engage in — vs. the idealization of formalized business processes.”
Moran has been working on this kind of thing since he developed Notecards at
Xerox PARC in the ‘80s; it was an attempt to support “informal work practices and
let analysts organize information in a natural yet structured way. There’s always a

DECEMBER 2005 RELEASE 1.0

UNIFIED ACTIVITY MANAGEMENT

The default mode of UAM is to support flexible and open a natural complement to formal workflow and B2B
interactions between people. Some interactions need processes. A workflow does not need to “micro-manage”
more structure; and we are exploring extensions, such as people; it can delegate a complex social activity to people
access control policies and constraint handling, in order to through UAM, which can help the people coordinate their
interoperate with formal processes, such as project plan- work, collect the relevant materials, pass the results back
ning and workflows, without turning UAM into another to the workflow, and produce an audit trail of their work to
workflow language. The philosophy of UAM is to rely on satisfy compliance requirements.

social interaction mechanisms to control activities and to From a paper shortly to be published in Communications
impose only as much constraint as is needed. Thus UAM is of the ACM, by three IBM researchers.

10

RELEASE 1.0

delicate balance between being too rigid — automating some work but turning peo-
ple off — and too loose — not providing any help to organize the work.”

In short, the idea is to give users tools to manage not just their information — such as
search, contacts databases and calendars — but to manage their activities. An activity
may be a defined project with an end (or a deadline), or it can be a continuing
process, such as managing a particular sales team (or even a particular person).
Consider what the spreadsheet did for financial modeling; IBM is essentially trying
to create an equivalent tool for project management — changing it from a main-
frame-style application into something flexible, ad-hoc, user-friendly and malleable.

At the When 2.0 workshop Moran’s colleague at IBM Almaden Research, senior soft-
ware engineer and lead “activities” architect Stephen Farrell, will be demoing Wax, a
prototype service which provides a server for representing shared activity descrip-
tions, a Web-based user interface for viewing and manipulating activity descrip-
tions, and Web-based API’s for using the service.

Some of the UAM work is moving out of research into products. The Almaden/
Cambridge Research UAM group is working with the IBM WPLC (Workplace,
Portal and Collaboration Software) division, which includes the Lotus brand. The
first activity-based product is called Activity Explorer and runs on the IBM
Workplace platform: UAM has added an Activity Pattern plug-in to Activity
Explorer that will come out early next year and will be shown at Lotusphere in
January. IBM is also working on incorporating many of the UAM ideas to provide
activity support across a wide range of products using open standards which can
work as a back-end to other services in the Web 2.0 world. Call it IBM 2.0.

Eight days a week

In essence, unified activity management creates a representation of activity itself as a
“first-class computational object,” which is a specific but flexibly structured collec-
tion of metadata that links associated people (with contact information), phone

WWW.RELEASE1-0.COM

calls, calendar items, documents, e-mails and the like. That flurry of e-mails, calls,
documents and people actually has an identity and can have an arbitrary internal
structure. It can be referred to, copied and modified with new participants and doc-
uments, and considered open or closed. People can be part of it (or not), with speci-
fied permissions, roles and responsibilities (i.e. documents or sub-activities that they
are responsible for).

For example, the activity of preparing a workshop includes (sub)activities such as
inviting speakers, writing an accompanying newsletter and organizing meeting
space. The overall skeleton of the activity is the same for all workshops, but the
specifics vary for each workshop. Some sub-activities occur some times but not oth-
ers, such as organizing a field trip, renting a separate space for a cocktail party or hir-
ing a performing elephant. The specific objects used by the activity — such as a
customer list — can be shared and sub-setted, used by multiple activities, etc.

The beauty of this approach is that it recognizes (unlike a project management tool)
that elapsed time or schedules are not necessarily as important as the articulation of
different sub-activities comprising the project, their statuses, participants, and other
facets of these sub-activities. Using a variety of interfaces and visualizations, user can
query for all open items, can see who has most tasks to complete, etc. It’s the equiva-
lent of being able to find e-mails that need an answer, but across a much broader
range of objects than just e-mails. People, documents or other items can be flagged
for follow-up, forwarded to others to handle, and the like.

Beyond that, the idea is to develop a standard, generic (“unified”) representation of
an activity and services to support this representation for all comers. The service has
standard Web APIs. People can build tools and applications that put, retrieve, and
manipulate activity information from this service. Moran says, “We encourage plug-
ins to other tools that interface with the activity service, so you can manage your
activities from whatever context or environment works for you. For example, we have
a plug-in to the Firefox Thunderbird e-mail system that connects to UAM. It allows
you to manage your e-mail as activities, not just folders. A new e-mail comes in; the
plug-in queries the activity service for possibly relevant activities and a list appears;
you then plop it into the right activity. Or you create a new activity.”

| know what we're doing this summer

Likewise, UAM has a plug-in for Firefox that allows you to stick any Web page you
visit into an activity or to query what activities refer to this page. And anyone can do
it. An IBM summer intern connected a cell phone to UAM. “So now if I get aI get a

DECEMBER 2005 RELEASE 1.0

11

12

RELEASE 1.0

call from you, a query goes to the activity service for the activities we're jointly
involved in, and displays them on the phone screen,” says Moran.

But this is not just for “personal” applications. One notion being discussed within
IBM is linking to a business modeling tool used by IBM Global Services. “Modeling
is a complex collaborative activity between analysts, technologists, and the customer
being modeled,” says Moran. “Of course, there are also IBM methodologies for doing
modeling — which are often worked around because they are seen more as overhead
than help. So the idea is to embed a UAM interface in one panel of the tool. This
allows the participants to manage their collaborative activity of modeling. Also,
methodologies can be presented as activity patterns. These are more likely to be
adopted because they directly help you get started on the modeling activity rather
than staring at a blank slate.”

Adds Moran, “We believe that activity management needs are everywhere, and we
aspire to make our activity management services a ubiquitous mash-up like Google
Maps. For example, Craigslist apartment listings can not only be mapped out, but
apartment-hunting activities should be able to be managed. If such activity patterns
are openly available, then people can learn from each other’s hunting techniques and
experiences.” He adds, “This kind of vision makes sense for IBM, where our real
strength is the ability to do integration. I would never have conceived of this at
[Xerox] PARC.”

Trumba: Shared lives

Trumba, by contrast, is a commercial venture, founded by Visio co-founder and
CEO Jeremy Jaech and a team of about 27. The initial business model is to sell a
Web-based calendar-consolidation and publishing service (OneCalendar) to group
organizers and other time-middlemen, such as sports teams, schools, churches, clubs
and other groups that organize meetings and events for their members. Then,
Trumba hopes, some of those members will realize that they too are group organiz-
ers or “schedule sharers” — for family members and other partners — and subscribe to
Trumba for their own use as calendar consolidators. Somewhat to the company’s
surprise (especially since the pitch is sharing calendars from the server), many indi-
vidual users are appearing, no doubt attracted by its ability to overcome a basic chal-
lenge — merging data from one calendar to another. On the other hand, says Jaech,
he’s surprised how hard it is to get people even to enter an e-mail address for a free
trial in this world of spam and mistrust. The company won't say how many users it
has, but it says it has 15 million events in its database that appear on users’ calendars.

WWW.RELEASE1-0.COM

Trumba’s Mix-In feature lets individuals view any combination of calendars using
the OneCalendar server; it doesn’t automatically find openings or conflicts but allows
users to visualize their schedules across multiple calendars. (A new release currently
in the works will add conflict-resolution capability.) Users can also e-mail entire cal-
endars or specific calendar entries — as invitations or just for information — to others
with Event Actions, a service that can translate among calendar formats including
Outlook, Yahoo! Calendar, MSN Calendar, iCal and any CalDAV-compatible calen-
dar including Trumba’s own OneCalendar. rOneCalendar shows you the sum of all
the schedules you select; you can selectively mix them into the calendar view you
want to see. Event aggregators such as newspapers or ticketing sites can link to their
calendars at the Trumba site, or embed a Trumba-hosted calendar into their own site.

Trumba’s business model allows anyone to see and save (events from) Trumba calen-
dars for free. However, a user who shares his calendars with others on Trumba’s web-
site or who uses Trumba’s hosted calendar service must pay a yearly charge of $39.95
(after a two-month free trial). Says Jaech: “The benefit we offer to event communica-
tors and event publishers is the communication and promotion of events to their

audiences. We think this benefit is worth paying for. The benefit we
offer to schedule sharers is Exchange-like calendaring tools without

the need for IT support. We think that is worth paying for, too.” TRUMBA INFO

Headquarters: Seattle, WA
He adds, “The benefit of using OneCalendar as a personal calendar Founded: October 2003
is having a consolidated view of multiple calendars. We don’t think Employees: 27

that is compelling enough to charge money for, and we figure all the DGS Sle eel eee

š è 3 Perkins, August Capital and Oak
personal calendars already in the market are going to add this func-
Investment Partners

eae ae rae
tionality anyway.” With this in mind, the company will release a free Key:metrie about i5:imitior events in tès

client early in 2006. database

URL: www.trumba.com

Right now, Trumba has rudimentary privacy mechanisms. You can
share a calendar for someone else to see or modify, or keep it pri-

vate, but that’s about it. Of course, you can create multiple calendars of private or
public meetings and share each of them selectively. Inelegant, but it does the trick.

The event fields in Trumba’s database are basically text strings, which can be copied
or set as recurring events for, say, every Tuesday, every five days or other periods a
user can specify. However, other than selecting recipients of reminders or invitations
from one’s address book, Trumba doesn’t offer much in the way of integration with
contact lists. Some might consider this a disadvantage, but for the broad market
Trumba is trying to reach, it’s probably about right — at least for now, while people

DECEMBER 2005 RELEASE 1.0 13

14

RELEASE 1.0

are new to the intricacies of sharing calendars. Although Trumba does have ambi-
tious plans — including “an open, event-router service that is interoperable with the
leading calendar services and applications, along with invitations, updates, notifica-
tions and reminders,” the challenge is not more features; it’s getting users to take
advantage of the basic capabilities in the first place.

Airena’s AirSet: Precision planning for the complicated life

On the calendar side, Trumba and AirSet are the two most directly competitive.
Trumba is easier to use, while AirSet wins on functionality. ..and also as a broader
group-support tool (assuming you want that). AirSet’s name and positioning focus
on its ability to operate on mobile phones — a specialty for Dougherty, who founded
and ran Geoworks, the company responsible for graphical operating systems and
applications for the Apple II, Commodore 64, and IBM PC, as well as mobile phones
and PDAs including Nokia’s first Smartphone, the Nokia 9000. That’s also key to
AirSet’s future business model, which is to offer the basic service for free, but to
charge for access to it over the air. Because its aim is to use each user as a viral mar-
keter to other members of all his various groups, that mostly-free business model is
important. Thus, in addition to calendar functions, AirSet offers more support for
contact management — another function that’s really handy when you're out in the
field or on the road trying to get hold of someone from your Treo.

What AirSet also brings to the party is granularity and natural support for the over-
lapping groups any single user may be a member of. It treats the participants in a cal-
endar entry not as text strings, but as participants, and can manipulate appointments
and show calendar subsets on that basis. That is, Juan can see all of Alice’s events; he
can also choose to see only those to which he is invited or which he has chosen to
attend. (In the future, he’ll also be able to filter his own appointments by which ones
include Alice or other third parties; this will be part of a grid view where you can see
events by user participation.) Currently, the “personal calendar” tab filters in all
events you re participating in, from all groups/calendars that use the AirSet back-end.
That granular filtering will enable a user to selectively consolidate schedules auto-
matically. Adds Dougherty: “It is configurable so that you can synch all your AirSet
data into Outlook [or a CalDAV calendar] or just do a one-way pull of Outlook data
into AirSet.” You may not want to add family events into Outlook, but you may want
to pull work events into AirSet to explain to your family why you are never home. ..or
perhaps you'll see the error of your ways and cancel some appointments!

WWW.RELEASE1-0.COM

While IBM and to some extent Chandler use “activities” as a way for
users to handle the complexity of all their disparate but overlapping
obligations, AirSet does so through group functions, which also
include shared to-do lists and (its own) blogs. AirSet has an overall
P2P flavor: A particular group’s contact list and calendar is shared
by default, and any member can update it. That devolves responsi-
bility. ..and control.

AIRENA INFO

Headquarters: Pleasant Hill, CA

Founded: June 2003

Employees: 14

Funding: $2.5 million from founders

Key metric: manages groups, not just
schedules

URL: www.airset.com

AirSet is likely to appeal to a more business-oriented user — one who
is probably using Outlook and may be loath to switch. That’s why the
cell-phone support — including an easy way to subset events to ones that matter in
real time — is so key.

Zimbra: Total control by owners

Zimbra, an open-source project, is complementary to Chandler’s open-source mes-
saging client, though it competes with the Cosmo server to some extent. Since so
many of our mail transactions concern our time (the financial ones we have already
made more explicit, with purchase orders and the like), Zimbra necessarily has a
heavy calendar component. “People spend their lives in Outlook,” says founder
Satish Dharmaraj, “but it’s broken. We’re an enterprise company, but our pitch isn’t
total cost of ownership or e-mail on Linux; it’s innovation in administration of the
server and innovation in end-user functionality.”

Zimbra comprises a mail/messaging server that attempts to detect the implicit trans-
actions within mail and provides hooks for corporate developers to do interesting
things from within mail — everything from seeing a map for an appointment, to
flowing a purchase order into an inventory system or flagging a request into a fol-
low-up folder or to-do list. That’s the appeal of open source; it invites users to define
and automate their own workflows.

In short, Zimbra wants to offer Total Control for Owners — letting people stay inside
the familiar environment of mail, but using it both as a dashboard to personal appli-
cations/data such as address books and calendars, and as a kind of client-side mash-
up between a user’s own desktop applications, plus specified corporate applications
ranging from Salesforce.com to an SAP inventory management system. (Of course,
it also offers better searching, foldering, tagging and the like.)

DECEMBER 2005

15

Dharmaraj was previously VP of Messaging Products at Openwave Systems, respon-
sible for the e-mail, unified messaging, voicemail, and instant messaging product
lines. But he’s just one of a team of about a dozen developer/managers who have
worked together since they worked together on Javasoft at Sun Microsystems in the
mid-‘80s. During the days when e-mail on the Web became popular, the team decid-
ed that it was time to bring voicemail and fax to the Web as well — so they moved on
to start Onebox, a unified messaging company that was sold to Phone.com for $850
million in 2000. Phone.com subsequently merged with Software.com to become
Openwave in 2001. At Openwave, the team ultimately deployed 180 million mail-
boxes worldwide, including 20 million mailboxes at JPhone and KDDI in Japan.
That record made it fairly easy for the new start-up to raise $4 mil-
lion from Benchmark and Redpoint in early 2004, with another $12

ZIMBRA INFO

million from Benchmark, Redpoint and Accel Partners together this

Headquarters: San Mateo, CA

year. (Only one key figure, president and CTO Scott Dietzen, was

ouaded De amber 200 not part of the persistent Javasoft team. He joined Zimbra from

Employees: 40

Funding: $16 million from

IBM, WebLogic and BEA.)

Benchmark,

Redpoint Ventures and Accel Capital Because it’s open-source, Zimbra is designed to be extensible and to

Key metric: five figures of server down-

loads in first 6 weeks

URL: www.zimbra.com

allow internal business managers or IT departments to customize
the applications and “recognizers” (to detect, say, a purchase order,
an expense item or a for-review request). The business model is to

16 RELEASE 1.0

offer the server free to all comers, and then to charge for customers

who ask for 24x7 support or for a variety of “enterprise” features
such as disaster recovery, replication, cross-mailbox search for legal discovery and
compliance-related functionality on the back end, and “over-the-air” sync on the
end-user functionality side. “We expect to charge about $30 per mailbox per year,”
says Dharmaraj. “We won't do it the traditional way, with a northeast regional man-
ager of sales in New York [or more likely Montclair, NJ]. Our customers will self-
qualify. If they’re smart enough, they'll download it for themselves and never bother
us. But if they want more, they’ll call us for support and extensions, and we'll sell
them add-ons and special features.”

This model might not have worked even a few years ago, but in a world MySQL,
SugarCRM and JBoss, it may be the first thing to offer a credible alternative to
Exchange. (Ironically, Zimbra won't get most of the revenues Microsoft stands to
lose, but it will erode the overall value of that market.) Zimbra has had several thou-
sands of downloads in the last three weeks. The company has 36 employees in the
US and 10 in Pune, India. It has also attracted a far-flung developer community that
has already created local-language versions for Russia, Romania and Brazil.

WWW.RELEASE1-0.COM

So, what does Zimbra actually do as far as time and scheduling are concerned?
Mostly, it detects references to specific times, for which it can open calendar entries,
and contacts, for which it links to (or creates) the appropriate address book entries.
(Shades of Anagram; SEE RELEASE 1.0, JUNE 2004.) That enables mail to operate natu-
rally as an interface to corporate data. It can do this for a variety of mail clients,
including not just Outlook but also Eudora (hurrah!), and with calendars including
iCal and therefore Trumba, Chandler and Sunbird.

Zimbra does not offer a fat calendar client itself, but relies on its Web-based AJAX
client for end-user functionality. It expects users to choose a desktop client for them-
selves, but if asked, Dharmaraj recommends Chandler. . .or will, once it is fit for
broad use. He adds, “The Zimbra AJAX client allows overlays of different personal
calendars as well as overlays of calendars of other end-users who have given explicit
permission to do so. It also allows subscription to any of the iCal-based calendars on
iCalshare.com, the calendar-sharing site. Zimbra also allows users to publish their
calendars for others to subscribe to, via a website or an RSS feed.”

However, Zimbra doesn’t do “activity management” in the sense of “knowing” about
activities. It doesn’t actually hold links between, say, a calendar item and a related e-
mail. It can simply find them through a powerful search capability. Thus, you can’t
associate purchase orders with an activity as in, say, IBM’s UAM model or even
Chandler; you have to reference them within a calendar item or e-mail, and then the
system can find them for you quickly.

Scheduling: The Transaction Side of Calendars

Calendars are really just a fancy kind of structured data store. The better ones have a
variety of capabilities for data reconciliation and display, but in the end they are
repositories for meetings scheduled by users.

How those meetings get scheduled is another, orthogonal business of social transac-
tions, and one that is heating up as more people schedule more meetings. Of course,
you can simply compare schedules — by hand or with the shared-calendar tools
described above. In a more automated approach, many business people use Outlook
to schedule meetings within an enterprise or other shared-server community, or to
send invitations outside it. But people familiar with such capabilities want a way to
use them outside their own corporate walled campuses.

DECEMBER 2005

RELEASE 1.0

17

18

RELEASE 1.0

One benefit of scheduling tools as opposed to shared—calendar tools is that they
don’t contain much specific calendar data, which helps to maintain privacy. The
users expose only the times they want to, according to the particular event and par-
ticipants. But that leaves most of the work up to the participants.

The leading “public” scheduling service is Evite, the most-used (and probably there-
fore most disliked) of them all. It was launched in 1998 and sold to Ticketmaster in
2001, and is now part of Barry Diller’s Interactive Corp. It boasts impressive month-
ly stats: 6 million unique users, 100 million page views, 250,000 events created and 8
million invitations sent.

Evite is basically a unilateral invitation tool: The event organizer specifies the event
and invites the prospective attendees. Evite collects the responses. It also adds value
by allowing recipients to post responses and answer polls — either for others to see or
only for the inviter. Registered users can store all their events and see related data. It’s
a solid old tool, but inflexible and ridden with incongruities. For example, it allows
recipients to specify which times they can or cannot attend a meeting, but they can’t
specify the same for a location. (Don’t we all know people for whom we would travel
across town — and others for whom we would not?) In the end, one gets the sense
that the main purpose of all this is to provide content around which to display ads.

There are a number of newer, more flexible scheduling services. The most capable
appears to be TimeBridge, but it is not yet even at the beta stage. Other, lighter-
weight entries include Renkoo and Skobee (who will be present but not formally
presenting at the When 2.0 workshop).

Renkoo: My time or yours?

Renkoo was originally designed as a lightweight personal scheduling tool. It started
its private alpha in November. For now, it has limited functionality, but it exempli-
fies many of the opportunities and challenges around scheduling.

While Renkoo performs functions similar to Evite’s, it is totally different in practice,
focused on negotiating and scheduling events that don’t already exist rather than
inviting people to ones that do, and then allowing users to import the appointments
they make into their calendars. While the average Evite goes to about 40 people,
Renkoo expects its users to focus on “lightweight” gatherings averaging closer to five
or six people. (Certainly, you wouldn’t want to try to negotiate a single time or venue
with more than that number, although many people do!)

WWW.RELEASE1-0.COM

In addition, Renkoo provides more group context: It can maintain a personal/shared
list of user-entered (or selected from ads) favorite venues and meeting templates —
i.e. the same bridge group as last time, but at a different person’s house, or the same
restaurant as last time, but with a different group of friends. ...

Also, Renkoo wants to avoid pushing people to the site. Ambitiously, it hopes to pull
people to the site, by offering more functionality and more information. “Our goal is
to let each user have the Internet experience he or she wants to have — whether that be
coming to a website, living in Outlook or getting real-time notifications via SMS,”
says co-founder Adam Rifkin, who previously co-founded KnowNow, a Kleiner
Perkins-funded middleware company, in 1999. Co-founder and CTO Joyce Park is
most famous for being fired from Friendster for blogging, but she is
also the author of “The PHP5 and MySQL Bible.” She was a lead

engineer at Friendster and also at KnowNow, where she met Rifkin. RENKOO INFO

They have worked together on a variety of projects since. ficsdaaadereeale Alto, CA

Founded: March 2005
Over time, Renkoo sees its business model as local advertising rather Employees: 5

than the brand advertising Evite features. Most interesting to local Funding: none; currently raising angel

advertisers, Renkoo will get to users while they are still negotiating roana ag 9y S98 Kraus

è Key metric: currently in private alpha;

on where to meet, eat and spend money. (Of course, Evite and
will launch in early 2006

Yahoo! Groups may pick up some new tricks. . .) tisha Sasheshinoaate

The most novel and useful part of Renkoo, not yet implemented but

promised for When 2.0 in December, is its voting capability, which will let users
negotiate on times and eventually places. “Everyone said they wanted Outlook inte-
gration [which is also promised for later],” says Park. “But when we mentioned vot-
ing, even though they hadn’t asked for it, they got much more excited.” That’s ironic,

since Evite already has voting. . .but most users don’t seem to know about it.
The Renkoo service currently has a somewhat cutesy, consumery-flavor. It may end

up with two versions, or it simply may include more business options such as meet-
ings and interviews as well as parties and movies.

TimeBridge: Game theory for meetings
You are invited to try out TimeBridge — but not until next quarter. No negotiating!

Back in 1997, Yori Nelken founded Banter, a company that used AI (statistical mod-
eling, natural language processing and machine learning) to help automate call cen-

DECEMBER 2005 RELEASE 1.0

ters and improve agents’ productivity. Now the Banter technology is doing well as
part of IBM, and Nelken is tackling the wider and much more common productivity
problem of scheduling meetings. His new company, TimeBridge, is still under
wraps, but he’s prepared to talk about the problems without disclosing the solutions.

After seven years at Banter, including two stints as CEO and two as CTO, he took
some time off to think about other interesting problems and settled on the challenge
of scheduling; it intrigued him that this common problem attracted so little atten-
tion and that no technology seemed to be available to support the process of getting
a group of busy people together.

Take the simple problem of us arranging our two initial meetings with him: Should
it be in Palo Alto or San Francisco? The first meeting was in San Francisco; the sec-
ond in Palo Alto. Should/could co-founder and VP marketing Bill Glazier, formerly
of PlaceWare (now Microsoft, page 23), join us? Yes the first time and no the second.
We don’t know what alternatives Nelken had, but we were trying to
meet with a number of calendar-oriented companies over the

TIMEBRIDGE INFO

course of several days — some in San Francisco and some on the

Headquarters: Menlo Park, CA

Founded: May 2005
Employees: 12

Peninsula. Were we busy at the times suggested? Yes, we already had
lunch plans the first time — but we changed them. As for the second
meeting, Nelken opted for a time slot we proposed, but by the time

Funding: undisclosed amount from he confirmed, that slot was not available. Back to square one.

Mayfield Fund

Key metric: launching next quarter

URL: www.timebridge.com

The lesson: Most scheduling involves other people and external con-

straints — the schedules of people more important, outside events

20 RELEASE 1.0

such as conferences, airline schedules, financial quarters, availability
of resources. Moreover, there’s an invisible hierarchy to the constraints, and it is all
very, very dynamic.

In fact, says Nelken, “People frequently make scheduling decisions on the basis of
who’s asking and who else will be there...and only then do they look at their own
schedule to see whether they are busy. For the right meeting, they'll cancel what
they’re already scheduled to do.” So for everyone except for the inviter, the schedule
is only one consideration — and not necessarily the first one.

Scheduling meetings is a negotiation, not an invitation, says Nelken. However, adds
company advisor Mark Drummond, founder of TimeDance (circa 2001), “Playing

calendrical battleship is a bad idea. You have to expose something — just not neces-

sarily your entire free-busy schedule.”

WWW.RELEASE1-0.COM

The challenge for TimeBridge is to figure out what to expose and how to handle the
responses, creating a de-facto ad-hoc workflow. A two-way negotiation is easy
enough, but suppose you want to have lunch with four people — and you are free
next Tuesday, Wednesday and Friday. Should you ask Juan first, because he matters
most? But then he'll pick a single day, even though he may be available all three days.
You could invite all four of them in a single message, but then the danger is that
someone else will pick a day that Juan, the most important person, can’t do it. (Or
you may want to have lunch with each of them separately, which creates a different
but similarly complex scheduling problem.)

Should you be transparent and let all of them pick — seeing which day Juan picked?
Youre taking the risk that Juan will come in last, with a date the others didn’t pick.
“Sometimes the pressure of the other three can get someone to change his schedule,
and sometimes not,’ notes Nelken. The company won't yet disclose its solution, but
Nelken leans towards transparency and openness as a default, in part because it fos-
ters better behavior. Nonetheless, the TimeBridge system will also allow double-blind
operation, where attendees specify their meeting desires and constraints, and the sys-
tem produces a match of free times without compromising anybody’s privacy.

We assume that TimeBridge, like so many Web 2.0 start-ups, will start free and charge
later for premium services “. . not inconsistent with Microsoft’s recent announce-
ments,’ says Nelken. He adds hopefully, “Once 100 HP salespeople are using it as
individuals, we'll call HP’s head of sales and offer the corporate, branded version.”

Also worth a mention is Skobee, another start-up. It addresses the same functions as
Renkoo, but it has a completely different take on privacy. Rather than hosting private
negotiations about where to meet, it posts all the information on the site, just as if
you were still living in the same dorm with all your friends, says co-founder Noam
Lovinsky, 25. For now, the system has three levels of visibility: Public, all my con-
tacts, and just my friends. The goal is to create a lively social space and let people
keep up with one another’s activities, check out where the gang is going Friday night,
or see who you can run into at the pizzeria before you head out. Advertisers, of
course, should love this.

Suppose you're shy? If Juan and Alice really don’t want anyone to know they’re see-
ing each other, yet they still want a meal out, they can each separately and publicly
arrange to meet with someone else at the same time and place. Then they just need
to figure out how to ditch their public partners before the evening is out.

DECEMBER 2005 RELEASE 1.0

21

22

RELEASE 1.0

Microsoft: Time of change

Ray Ozzie’s arrival at Microsoft as CTO both reflects and drives fundamental changes
at the company. His entire ethos — from flat organizations and peer-to-peer to a cer-
tain sensitivity to the human element — is a shift from Microsofts more traditional
command-and-control structure, yet without the pure Darwinism of the Google
approach. We expect to see the company paying more attention to human factors and
to collaboration, both in its products and in its development process.

This is all part of a new attitude at Microsoft. The power shift in the industry — with
Google taking both the glory and the barbs that were formerly Microsoft’s — seems to
have left Microsoft with new freedom and a willingness to change. We expect to see
the company experiment more. Microsoft Live, Microsoft’s move from the desktop to
the Web and to software as a service, is just the first of many new experiments.

Another innovation is peer-to-peer information feeds; that is the approach of the
Simple Sharing Extension, a recent Microsoft initiative led by Ozzie to address calen-
dar and other sharing issues people have today. Ozzie, of course, created Groove, the
secure, shared-workspace software designed for users working across corporate
boundaries. (SEE RELEASE 1.0, MARCH 2001.) Groove is now part of the Microsoft Office
system. Using feeds across borders is exactly the kind of approach Groove took, but
only within a closed circle.

We don’t know what Microsoft plans to do with it, but this Simple Sharing
Extension initiative is an interesting development — two-way feeds replicating
between servers (or between clients and servers) rather than typical one-way RSS
feeds. Rather than just editing calendars within the context of a single server or web-
site (e.g. Exchange or MSN), individuals or sites can share and co-edit sets of calen-
dar appointments (or other data) among such calendar stores directly through
Notes-style replication, via an RSS feed made bi-directional by a set of RSS 2.0
extensions. As with Notes (no surprise!), this replication could occur client-to-serv-
er, or server-to-server or site-to-site; they’re all peers. (If it weren't for firewalls, it
could even occur directly client-to-client, as it can through Groove.)

Feeds might be created or edited by the Outlook client, or by the Exchange server, or
by any website with calendaring capabilities. Any SSE-enabled feed can then swap
appointments or events with any other. Though this method doesn’t address multi-
user scheduling negotiations such as those handled by Exchange or TimeBridge (pace
19), the offsetting value is that users can share such information without being limited

WWW.RELEASE1-0.COM

to a community of people on a single calendaring server, and by extension they can
edit all their shared information in the calendaring UI of their choice. Stay tuned.

Negotiating changes: Bob and Carol and Ted and Alice

Sometimes time doesn’t matter but timing — or the sequence of events — does. That’s
behind another project at Microsoft, led by Pavel Curtis. Curtis started his career at
Xerox PARC, where he created LambdaMOO and was hailed for shaping that
decade’s vision of multi-user environments. He went on to co-found PlaceWare, a
Web conferencing service. He and his partners and investors sold it to Microsoft in
2003 in the recognition that ownership by Microsoft was likely the best way to get
their service into the hands of multitudes.

Curtis is still working on multi-user collaboration, but now he’s focused on the
intricacies of reconciling multiple users’ changes to a single shared work product
and obliterating the impact of timing — the sequence of the edits — on the outcome.
We use documents as an example here, but the approach addresses any kind of con-
tent, including spreadsheets, presentations and source code.

Working with colleagues at Xerox a decade ago, Curtis had already solved this prob-
lem once, but primarily for the special case of only two users. That solution scales to
more users only if you stick to a client/server architecture, where clients talk only to
the server and never directly to one another, as in PlaceWare/Microsoft’s Web con-
ferencing software (called Office Live Meeting).

The general problem, with multiple users all talking to one another peer-to-peer, is
much more difficult. “I had left this area behind [during PlaceWare],” says Curtis,
“but Pd heard rumors several times about it being solved in a number of different
ways. But when I came back to it, I discovered that things really werent much fur-
ther along. Mostly, we just understood better why the problem was hard, because so
many attempted solutions were proved wrong.”

The general goal is to let all the users work independently, as if each was the only one
editing the data. There are no locks, no turn-taking, no barriers to making progress,
even if other users are working at the same time.

The trick is to understand the intent of the edits, so that they can be transformed

appropriately for intermediate changes. For example, suppose one edit was “delete
the fourth character” and someone else has deleted the second character. The inde-

DECEMBER 2005 RELEASE 1.0

23

24

RELEASE 1.0

pendent edit needs to be transformed to preserve the original user’s intention, into
“delete the third character,” before it gets applied. In practice, of course, edits tend to
be much larger, but the idea is the same.

(In case you were wondering, no, the software doesn’t recognize paragraphs (unlike,
say, document-compare tools); it simply sees the document as one long string, says
Curtis. “Paragraphs don't really exist, per se; they’re just the sequence of characters
that happen to appear between two carriage-return characters.”)

This example is pretty simple, but it can get very complicated very quickly: What if
Alice has seen all of Bob’s edits but none of Carol’s when she receives new edits from
Ted, who has heard from Carol but not Bob? It’s hard to visualize all of the possible
interactions, let alone create software that can sort it all out.

Curtis figured out the math of precisely defining the low-level intent of individual
document changes and then preserving those intents across other changes. “It has to
look, in the end, like all of those intended edits were performed in the same order on
every user’s computer, even though different users saw them happen in different
orders.” The order may be arbitrary, but it must be consistent for all users. At the
end, the “owner” can see the net effect of all the changes, decide which ones to accept
and resolve any clashes of intent — either lower-level or at the “do we support evolu-
tion or intelligent design?” level.

“I don’t have a proof yet that these transformations are always correct,” says Curtis,
“but I wrote a program that simulates four users pounding away randomly on the
keyboard and ran it for about half a million edits. Their documents stayed in sync
the entire time, so the statistical evidence is pretty solid.”

Of course, the point isn’t to resolve all the conflicts; it’s to “remove” timing, so that
the order of the edits (which is unknowable in the disconnected situation described
here) doesn’t affect the outcome. Instead, it presents the results of the edits in a way
that preserves the editors’ original low-level intentions, and that lets the party in
control resolve the higher-level conflicts, either by himself or by checking back with
Ted (who has changed “evolution” to “intelligent design” throughout) and Alice
(who wants it changed back).

How will this be applied? We can see lots of uses for it within Word, for editors try-

ing to collect changes from colleagues — or from the sources and subjects covered in
a newsletter. Likewise, changes in business-modeling spreadsheets, presentations to

WWW.RELEASE1-0.COM

senior management, and software source code are frequently made collaboratively,
involving the synchronous or asynchronous participation of multiple people.
Today’s solutions to these situations are pretty clunky, time-wasting and difficult.
Most people resort to laboriously taking turns making edits or just appointing one
member of the team to be the only one who can make changes. With luck, Curtis’s
work will eventually make all of this a lot easier.

Currently, Curtis is part of Microsoft’s Real-Time Collaboration group, still involved
with Office Live Meeting. But his goal is eventually to support other groups as well,
including all the Office products.

Events on Parade - In Public Portals

At the other end of the spectrum from fine-grained scheduling tools are public event
databases, which hold a registry of events and allow you to download them selective-
ly. Most notable in the straightforward public-events space are Upcoming.org,
recently acquired by Yahoo!, EVDB/Eventful.com, and Zvents. Then there are other
takes on “events”: Columba focuses on market-moving events of interest to banks
and investors; Meetup helps create meetings bottom-up; EventWeb (ambitiously)
wants to index and annotate all the news about all events; and Demand ID wants to
collect demand for events and get musicians and others to schedule live events in
response to those demands. They all interact and intersect, both in concept and in
some cases, in the market.

Finally, we include Technorati, in the person of chief technologist Tantek Celik.
Technorati’s business of searching blogs and analyzing the dynamic blogosphere
depends on time and makes the movement in the blog world explicit and visible. But
Celik also wants to add explicitness at a lower level by making references to events
(and other social phenomena such as people or venues) concrete and precise, by fos-
tering the use of the hCal microformat to represent them. Then events portals can
continue to add value, but the “identity” of an event will be easier to represent and
more portable across the Web.

Indeed, in the absence of ubiquitous standards such as hCal, all events sites face a
variety of delicate business problems around how to get and define event content —
from asking users to do it all and scraping websites, to sucking in feeds and forming
actual business partnerships with a variety of event organizers and promoters.

DECEMBER 2005 RELEASE 1.0

25

26

RELEASE 1.0

Although this isn’t like downloading tunes or even providing an index to all the
world’s books, there are still some delicate IP issues to consider.

In a world where standards reign, it will be easier for anyone to share calendar infor-
mation either publicly or to defined lists of counterparts. Rather than publishing to

calendar portals such as EVDB, Zvents and Upcoming, event managers — along with
bloggers and random other folks — will be able to share calendar entries and (if they
wish) allow users to edit or create entries.

This means that event aggregators will need to do more than just operate as event
registries or search engines — something that their direct competitors and Google
already do quite effectively, thank you. Says EVDB’s Brian Dear, “Just because more
publishers or event managers are publishing straight to the Web — thanks, we hope, to
tools we have a hand in making — doesn’t mean that aggregators serve no useful pur-
pose anymore. On the contrary — their usefulness will flourish as we become awash in
a sea of events.” Over time the challenge will become not finding, but filtering.

So, the events portals need to become, well, a Flickr for events, where users can com-
ment about events, share them with their friends, point to them, tag them, rate them
and so forth. (Note that Flickr is now a sister company to Upcoming.org.) Certainly
there are differences. Users don’t “own” events the way users own photos. And event
creators are closer to advertisers than photographers in their motivations. Events are
also more complicated — in the sense of having moving parts — than a photo. They
may have venues, performers, participants, tickets, attendees. . .all fodder for a variety
of interesting business models and complex ancillary functions and services. And of
course they have formatted information that can be imported into a user’s personal
calendar, or sent to a third party as an invitation or notification. (Imagine Alice post-
ing an event link on Facebook with a request for someone to buy her a ticket. . .)

Same time, next year

Over time, event databases will become potentially powerful social spaces, where
users plan events (or at least plan to go to them), request events (SEE DEMAND ID, PAGE
36), discuss and recommend them, purchase related content or travel services or
invite their friends. After the fact, they can rate events, post photos (or link to photos
elsewhere), discuss what happened, and possibly plan follow-up activities.
Individuals would register in order to post, comment and rate events, which would
reduce spam and support a reputation system. They could also determine who could

see their own participation plans, though in general the number of intentions for

WWW.RELEASE1-0.COM

any particular event could be disclosed without individuals’ permission. There are
numerous social conventions that remain to be addressed.

Other business models include selling not just ads but market information back to
event organizers (e.g. Demand ID), and also to related parties: travel and hotel com-
panies, vendors of related content, and planners of all kinds.

Long-term, we can also imagine a spot market for individual performers. For exam-
ple, if Bruce Sterling happens to have a five-hour layover in Denver on his way from
Austin to Seattle to see Neal Stephenson, there’s certainly a critical mass of fans in
Denver that would be willing to haul out to the airport to hear him. For his part, he
could have some fun, sell some books and perhaps collect a fee or at least get a free
meal at the airport deli.

That’s if these companies do it right.

Overall, EVDB sees itself as a user-driven site, whereas Zvents is more advertiser-dri-
ven, focused on “the official story” and on reliability of information. Even though
Zvents offers roughly equivalent tools for individuals and groups to post events,
share calendars, add related content/comments and integrate with blogs and other
sites, EVDB sees each event as a blog with its own permalink, whereas Zvents sees it
more as a posting. Although the overall functionalities are similar, the EVDB
approach is likely to result in a site that is livelier, noisier and intrinsically more sus-
ceptible to spam, although active users are good at policing their own turf. . For its
part, Upcoming is totally a user-driven site, although it is likely to change character
somewhat under the influence of Yahoo!’s ownership and also Yahoo!’s users.
Lurking in the background somewhere is an offering from Google.

Upcoming.org/Yahoo! Local: Up and coming
Upcoming.org, an early innovator in the events/calendaring space, launched in

2003. It was the first site to syndicate events using RSS, and the first with a public
API for events. It was also one of the first sites to foster social networks around any-
thing other than meeting people.

“Tt stemmed from a problem I wanted fixed,” says founder Andy Baio. “I was losing
track of the events I knew about, and not finding out about other things until they
ended.“ He bought the domain name early in 2003 and launched the site in
September 2003, focusing on Los Angeles events but providing a platform for users

DECEMBER 2005 RELEASE 1.0

27

to add events globally. “I let the roads pave themselves,” he says. “It ran itself for a
good period of time, with minimal effort from me.” Users were the sole source of
events, which has kept the site’s database relatively small but has given it much more
of a community feeling than either EVDB or Zvents.

In fact, it was so casual that Baio kept his day job as webmaster at Dimensional Fund
Advisors until the Yahoo! acquisition this fall. However, he began taking Upcoming
more seriously back in March 2005, when tech columnist Jon Udell mentioned the
site in his Infoworld blog. He praised it for the most part, but wondered why it had-
nt caught on. Recalls Baio, “He proposed adding tagging and a pub-

UPCOMING.ORG INFO

lic API, which got me motivated to do more development. At that

Founded: 2003

URL: www.upcoming.org

Headquarters: Sunnyvale, CA

point, two friends of mine volunteered to help with the coding and
make something great.” Those friends, Gordon Luk and Leonard
Lin, are now listed as co-founders of Upcoming.org, but were never

Employees: 3 ‘ P :
f hired as developers — until Yahoo! “acquired” them.
Funding: acquired by Yahoo! in October
2005
Key metric: first kid on the block Unlike EVDB, which had investors and a profit motive, Upcoming

hadn’t focused much on growing as a business. When Yahoo!

28 RELEASE 1.0

showed up, says Baio, things clicked: “Yahoo! has a staggering
amount of resources and they're very interested in fostering innova-
tion. They were a perfect match for us, and the opportunity to develop our work
full-time and introduce it to a much wider audience is very exciting.”

Moreover, Baio feels that he can learn from Flickr in particular. Another recent
Yahoo! acquisition, Flickr focuses on the social interaction around photos; Baio
wants to focus on the social interaction around events. “Friends’ events are inherent-
ly more interesting than others’ events,” says Baio. “Yet you still want some serendipi-
ty. You want to be able to stumble upon things and have them surfaced for you the
way you do in Flickr, rather than search for something or get lost in Google images.
Right now, for example, I have over 10,000 items in my feed reader. Pd love to have
some interestingness measure” — akin to what Flickr offers, based on measuring vari-
ous kinds of user reactions including views, tags, favorites and comments, but dis-
counting friends’ assessments in favor of those of objective strangers.

...and from Yahoo!'s corner

Yahoo! has been active on the acquisition front lately, often going after people as
much as properties. Its most notable acquisition this year, Flickr, brought only a pal-
try number of users compared to Yahoo!’s own millions. The same is true of

WWW.RELEASE1-0.COM

Upcoming. And last year’s acquisition, Stata Labs and its mail tool Bloomba, has

gone quiet. (SEE RELEASE 1.0, JUNE 2004.)

But Stata Labs’ founder, Raymie Stata, is still with Yahoo! and now, as chief architect
of its Search and Marketplace business unit, he plays a key role in its technical strate-
gy. He sees time and events as central to Yahoo!’s mission to help people “find, use,
share and expand human knowledge.”

In the past, he says, “The industry took enterprise functionality in calendars and pre-
sented it to consumers and said, “You know that calendar stuff you hate so much in
the office? Now you get to use it at home. Lucky you!’ But recently there’s been inno-
vation driven by players like Upcoming who started from a consumer perspective.”

He sees these new consumer players focusing on four “buckets of functionality:” dis-
covery (for example, through search), logistics (such as setting up a meeting), com-
mitment management (keeping track of your own commitments, a part of IBM’s
larger ‘activity management’) and annotation (sharing content about the content).
“Older products — including Bloomba — were caught in a rut,” he says. “They focused
on logistics and commitment management, and did so within a rather narrow
design space. It’s exciting to see these new players like Upcoming inject new life and
new ideas into the mix. I expect to see ideas from these products — such as commu-
nity-based discovery and annotation — make their way into enterprise products.”

Stata’s take on time was reflected in Bloomba, which was notable for its strong search
capabilities as well as for functions such as automatic filtering and archiving of mail
after a defined interval. For example, if you didn’t process something within three
days, it would be discreetly tucked out of sight but still be easy to find. In fact, Stata
would tend to extend a mail point of view to many things, including search. He says,
“If you rank things by time, as most mail systems do implicitly, it is a strong predictor
of relevance. More speculative is allowing people to browse by time —as blogs do.”

Moreover, he adds, when users search their own information they tend to remember
things by time. That doesn’t mean necessarily by dates, he says: “ People remember
things off events. They'll say, ‘I remember where you told me that and when. . . . Ah,
now I remember what you told me? We had a patent at Stata Labs, which Yahoo! now
owns. The idea was to use events rather than formal time. Key events for various
time periods would pop up. . .either from your calendar, or public events such as
9/11. ..which would help you more quickly identify the desired time period.”

DECEMBER 2005

29

30

RELEASE 1.0

He adds, “In many of these ideas for using time, place is a part of the story as well.
For example, photos have time stamps in them [although the mistimed camera may
be almost as common as the blinking VCR]. Getting location as well would be
handy. At Yahoo! Labs in Berkeley, we are playing with joining the time stamps from
your camera with GPS data from your phone, so we can determine where you were
for each photo.”

EVDB: Eventfully yours

EVDB, like Upcoming.org, was one of those I-want-it-for-myself projects. Launched
at PC Forum last March, the company recently re-launched as a platform (EVDB)
supporting the site Eventful.com. (DISCLOSURE: ESTHER DYSON SITS ON THE BOARD.)
Founder Brian Dear bought the domain evdb.com and founded the company in
January 2004, but he had wanted to develop such a service since back in the “90s; at
that point, the vision was a personal notification service, but the issue of getting the
content and the lack of funding made the idea a non-starter. There was no ready
source of content for alerts such as the Web offers today, even though it is still mostly
unformatted and needs to be scraped and parsed rather than seamlessly “fed.” (see
pace 40.) And certainly there weren’t enough users (or tools) around for user-gener-
ated content.

Later on, as director of Web development at mp3.com in 1999, Dear worked on its
notify-me service, which alerted users when bands they had selected appeared in
their city. The music thread goes beyond the notion of musical events; Dear had
online CD registries CDDB (SEE RELEASE 1.0, SEPTEMBER 2003) and IMDB in mind
when he named EVDB. Like the CDDB database of CD listings and the IMDB movie
database, EVDB is a database of information about events. It doesn’t “contain” the
events themselves, but it links you back to the source site or data.

EVDB is busily adding — or attracting — value to Eventful.com, most notably in the
form of user-generated content and community features. Dear sees each event as a
high-level object akin to a blog, with its own identity, ping and trackback capabili-
ties. Users are invited to comment about events, share and point to them, cross-ref-
erence them and so forth. His goal is to foster community around the events as a way
of getting them to attend. . .whereas Zvents is more focused on reaching would-be
attendees in a more traditional publisher/reader model.

The Eventful site also uses a variety of tools and techniques to generate event entries,
from deals with third parties such as Meetup.com to crawling event-laden sites such

WWW.RELEASE1-0.COM

as Ticketmaster.com (with permission), to encouraging users and event-heavy orga-
nizations such as training companies and booksellers to add their own events.

Dear says he expects it will take until late 2006 or early 2007 to achieve what he calls
the “flip” — to get to 51 percent user-generated content. Currently the proportion is
less than 10 percent. “The ‘user’ in user-generated doesn’t mean individuals,” he
adds. “It could be organizations or other third parties. But the point
is that we don’t need relationships with them. It’s self-serve. They

can use our tools and don’t need personal attention.” The attention EVDB INFO

—and the spam-catching, one hopes, come from other users. Much Hëadguärtėrs: San DiegoCĂ

of the quality control, he adds, will happen within groups (coming Founded: January 2004

soon), self-selected groups of users who can set their own privacy Employees: 7

controls and filters. Funding: $2.4 million from Draper Fisher

One notable feature on Eventful is the SmartCalendar, a prospective
search tool that lets users “search the future” for events that have not
happened yet. EVDB added SmartCalendars as a response to failed

Jurvetson, Omidyar Network and

angels including Esther Dyson

URL: www.evdb.com

Key metric: 300,000 events in database

searches — i.e., when a search has no results. SmartCalendars gener-

ate RSS and iCalendar feeds, either of which a user can subscribe to.

At first the feeds may be empty, but as soon as one or more events match the original
search, the feeds are populated. A user can also choose to receive an e-mail alert
when matches occur. That keeps them coming back — presumably to find more rich-
ness each time they visit.

Zvents: Suddenly last summer

Zvents is an upstart, beginning with its snarky “delta” launch last month. Indeed, the
site is full-featured in a world of shameless alphas and persistent betas. The company
was incubated by Net Service Ventures, a consulting company that counts several
large non-US telcos among its clients. “This was the most interesting of the 12 ideas
we proposed to one of them,” says Ethan Stock, then a partner at NSV and now CEO
of Zvents. “We started thinking about what people would want on their phones and
what they could generate on them.” Zvents has no particular relationship with that
client, and it’s targeted squarely at the US market to start, but in the back of Stock’s
mind is the notion that some day a lot of traffic about events will be real-time — both
searching and photo-posting. For now, the site assumes a regular PC screen.

Stock uses traditional city guides and newspapers as his model; he sees his job as
aggregating and making sense of “a cacophony of disastrous formats” with the help

DECEMBER 2005

31

ZVENTS INFO

of a variety of scraping and parsing tools. The team of 10 includes a

Headquarters: Menlo Park, CA

Founded: March 2005
Employees: 10

Funding: undisclosed amount from

couple of natural-language experts along with editors with a human
touch — and only one marketer, he says.

Zvents itself uses a dynamic calendar interface to display group or

NetService Ventures, Red Rock personal calendars of user-selected events in day, week or month

Ventures and angels

Key metric: listings for 5000 events per
week in the Bay Area; 80,000 total

URL: www.zvents.com

32 RELEASE 1.0

modes, though the main response to a search is a list.

“We want to get people out and doing things,” says Stock. The chat-
ter about events is secondary. That’s partly a reflection of Stock’s
own point of view; it also resonates well with event promoters — as

opposed to people who just enjoy posting about events. So the site
already includes links to maps and a “when” search, with an inherent “action-orient-
ed focus on today or right now,” Stock points out.

He adds, “We do only real events, with a specific time and place — no broadcasts on
NPR, for example.” But he would consider an event in, say, Second Life (an online
virtual world managed by Linden Labs), at a specific time and virtual place, to quali-
fy. Another restriction is no invitation-only events; the events must be open to any-
one, even though pricy tickets are not a problem. So far the system has about 80,000
events listed, mostly in the Bay Area, where Stock aims to gain critical mass first.
“When you start digging around,” he adds, “you find a ferment of niches — singles
mixers in San Francisco, art weekends in the Hamptons, square dances for seniors or
swimming lessons for kids in the Bay Area.”

Users can have a personal calendar page. This is more a listing than a social page,
without (yet) much information about the individual’s plans — which my be just as
well since for now it is visible to anyone. Any user can also place a Zvents widget in a
blog or on a Web page; it can show either a single event or a dynamic calendar, which
shows a list of current events or a single-day view that can be scrolled within or from
day to day by readers of the site.

The distinctions between Eventful and Zvents are subtle: While Eventful focuses on
user-defined groups and calendars and a list interface more conducive to discussion
than scheduling, Zvents considers a vendor-sponsored calendar of, say, all Release
1.0 events to be a fine idea. The two companies are likely to become more different
over time, just as twins growing up together and competing are often more different
than twins growing up apart.

WWW.RELEASE1-0.COM

Columba Systems: Time and money

Some people might find all these personal events a little, well, frivolous. At the other
end of the spectrum is Columba Systems, a UK company reincarnated from a Reuters
dot-com experiment called Kalends (Latin for the beginning of the month) that cost
about £20 million (roughly $40 million). The beneficiaries of that expensive educa-
tion, which ended in early 2002, regrouped as Columba Systems later that year.

CEO Dean Ratcliffe had always been interested in the interplay between structured
data and the softer, less formal stuff of journalists’ reports. He helped build a Reuters
tool that would let reporters fill in a form and would then produce a formulaic story
such as an earnings announcement, in multiple languages. But the reporters would
never use it, he recalls; they preferred to “write creatively” in flat, routine sentences
that carried no more information than the forms. It’s this interplay between unique
information and standard representations that continues to intrigue Ratcliffe and to
underlie Columba Systems.

Kalends and then Columba (for pigeon, an oblique reference to Reuters and its bird-
carried-news heritage) reflected Ratcliffe’s realization that the Reuters journalists
were swimming in real-time information but had only spotty
knowledge of future events, a condition that also afflicted Reuters’

mostly financial-sector customers. COLUMBA SYSTEMS INFO
Columba now has 18 people (vs. 52 for Kalends at its peak), and is Founded: April 2002

focused strictly on the paid-subscriber market of customers in Employees: 18

financial services. Its event structures are fairly routine: mostly earn- Bunding £400,009 fromangels nee:

ings announcements, public forecasts, trading statements, analyst
calls, investment conferences, trade shows, private board meetings
[reading the minutes is not the same as being there!] and public

official statements of various kinds ranging from financial statistics

ing John O'Connell, former CEO &

chairman of Staffware

Headquarters: London, United Kingdom

Key metric: 10 large customers; 35,000
events (10,000 active)
annual meetings, as well as public holidays, bond maturities, and URL: www.columbasystems.com

to budget forecasts. These aren’t necessarily events you can attend,

but they are events that can move markets. And now, with Columba’s aggregated
data, it’s also easier to detect historical patterns, looking at, say, oil or auto stock
movements following OPEC meetings over the past 10 years, or, more mundanely,
bond yield movements following the publication of labor-market statistics.

These are not secret events, notes Ratcliffe, but in the past they were difficult. . not

so much to predict as to find. Any transportation analyst, for example, could tell you
when FedEx was scheduled to announce earnings, but the information wasn’t aggre-

DECEMBER 2005

33

34

RELEASE 1.0

gated. All the analysts in a bank would fill in their own spreadsheets and send them
to the head of research; they would be consolidated and sent back out, but that rarely
happened before the events themselves had occurred. Now, Columba lets customers
consolidate both their own private events and Columba’s public ones, and offer that
to their clients as a service. In the future, we expect Columba will move beyond let-
ting people know when events occur, to providing data around events so as to make
other events — or price movements — easier to analyze and predict.

Columba currently has ten customers; each of the top 10 City of London investment
bankers is either paying for the service already or using it on a trial basis. For now,
pricing is negotiated with each client on the basis of seats, usage and the like; the
company expects revenues of about £500,000 this year (2005) and “significant
growth“ next year, says Ratcliffe. Meanwhile, he points to a counterpart, StreetEvents
(originally called CCBN for Corporate Communications Broadcast Network), a
subsidiary of Thomson Financial. StreetEvents has a slightly different, more content-
rich model, but it’s worth noting that Thomson paid $175 million for it two years
ago. Conclusion: Reuters doesn’t get timing. Let’s hope that Columba does!

Seraja, the EventWeb: A million single points in time and space

Perhaps the most ambitious events project we have come across is EventWeb, the
brainchild of Ramesh Jain, now Donald Bren Professor in Information and
Computer Sciences at University of California, Irvine. Prior to this he was a titled
professor at Georgia Tech.

But he’s not just a theorist. His background as a successful serial entrepreneur gives
some credibility to his undertaking: Born in India and educated at Nagpur
University and the Indian Institute of Technology at Kharagpur, he started the AI
labs at the University of Michigan in 1986. During that time, he founded a company
to work on the reverse-engineering of mechanical parts commonly used in automo-
bile, aerospace and entertainment industries — an intriguing notion. You aren’t just
reverse-engineering a product, but trying to reverse-engineer the constraints and
design considerations — i.e. the intelligence behind the design. He sold that company,
called ImageWare, to SDRC in 1998.

He then went on to found another university lab, this time the Visual Computing
Lab at UC San Diego in 1993. That same year he started Virage to develop its Visual
Information Retrieval engine, which could detect objects in video streams and clas-
sify the videos. Autonomy, the previously text-only taxonomy and classification

WWW.RELEASE1-0.COM

company, eventually bought Virage in 2003. And finally in 1996, Jain co-founded
Praja, which was designed to pick out the magic moments in sports videos; i.e.,
which sequence is a touchdown versus just the usual scramble of football players
charging into one another? That company was before its time, and in 2002 it was
sold to Tibco, which uses the underlying event application to monitor business
processes for specific events and generate alerts, security warnings and the like.

Now, Jain is taking all that he has learned over the years — and all that the Web has
created over the years to support such ambitions — to build what he calls EventWeb.
The vehicle is a company called Seraja (which has no meaning, but did have a URL
available). He is chairman; the company’s CEO, Arun Katiyar, is based in Bangalore
along with the development team; the primary investor, Rajesh Jain (no relation), is
also based in India. “Were moving from the document Web to the event Web,” he
says with the smoothness of a popular professor. “We want to move beyond the
Gutenberg legacy and reverse-engineer the environment. You should be able to fol-
low your own course through an event online.”

So far, Jain points out, most calendars are devoted to planning. He wants to use the
calendar as a high-level index and create something he compares to Pensieve in the
Harry Potter books: “You take out someone’s memory, put it into

Pensieve and everyone can share it.”
SERAJA INFO

The idea is to index and display content by time and place — i.e. to Headquarters: Bangalore; India
index events. And then — here’s the magic — EventWeb will process R E ea
the content it finds or gets from users using the sorts of pattern- and

team
l pR ; Z; l
object-recognition tools that characterize much of Jain’s previous Finding: Similien fromanesiihe
work. What makes it interesting is that it will can process video founder
objects as well as text-based event information. The service relies on Key metric: still an early-stage project
indexing, classification and recognition algorithms. . .and people. As URL: www.seraja.com

Employees: 10, plus outsourced technical

a service, it will both host its own content and object recognition,
annotation and editing tools, and let users use the tools to manage
and host both shared and their own content, with links to EventWeb. Imagine
Wikipedia-style collaboration to generate metadata for any event-related content
anyone can find.

The tools will do much of the work, and then humans (mostly paid contributors in
Bangalore — at least at the start) do the rest. For example (for now), while it might
take a human to distinguish the individual players on the field, the system could rec-
ognize a winning hit, strike-out or other sports action. The company will launch its

DECEMBER 2005 RELEASE 1.0

35

36

RELEASE 1.0

service first in India, with a focus on Bangalore and ever-popular sports events, as a
test market to iron out the kinks and gain some critical mass before, he hopes, users
start driving the system with their own content and metadata contributions. That
gives it a natural, highly involved audience, an initially constrained content space
and a skilled workforce to do the human tagging and annotation.

EventWeb is designed both for public events, and as a way for individuals to store
and manage (with a variety of tools) their own memories — for a subscription fee. “I
know my daughter got married,” says Jain. “I don’t need a calendar to remind me.
But I enjoy watching the video again and again, and picking out [and tagging] my
favorite parts.”

The challenge, of course, is finding the event content in a world where content is not
generally geocoded or time-stamped (including time-stamps that refer to the con-
tent, rather than when it was created, so that, for example, both Juan’s pre-birthday
hints as to what he would like and his post-birthday thank-you letters are time-
stamped to the date of his birthday). That’s why, of course, EventWeb is starting in a
fairly constricted space with an audience of likely-to-be-active users who can add
tags and structure to the data.

This is still an early-stage project, but it makes use of most of the technologies Jain
has developed in his long and productive life.

Demand ID Systems: If you demand it, he will come

Demand ID Systems turns the traditional model around: Instead of looking for what
scheduled event exists, users ask for it to happen. Specifically, instead of live-music-
events producers marketing to customers, customers express their interest to artists
and event organizers. Live music is an amazingly inefficient market, says Demand ID
founder Ben Cruz. He cites executives of Ticketmaster and Clear Channel Enter-
tainment, who told him that half the available audience capacity of music venues
went unsold — despite $1.5 billion in direct advertising and another $1.3 billion in
sponsorship funding (to generate revenues of about $10 billion in the US). And that’s
only for those venues during a performance; many of them are ”dark” several nights a
week. Compare that to the airlines, which currently have load factors around 70 to 80
percent. . .and fly most of their aircraft every day. Clearly this part of the music busi-
ness, as much as the copyrighted-content side, needs some rationalization.

WWW.RELEASE1-0.COM

The company was started three years ago, but mostly as a part-time,
underfunded project by Cruz, a longtime marketing and technology
executive. After 11 years managing international marketing, licens-
ing, and product development for companies including Black &
Decker and Noise Cancellation Technologies, Cruz founded
Synchromedia, a dot-com city guide that was among the most pop-
ular information services on Motorola’s ill-fated iKno service.
Neither service lasted through the end of 2001.

Demand ID is a different take on things. The idea is to get users to

DEMAND ID SYSTEMS INFO

Headquarters: Baltimore, MD

Founded: January 2003

Employees: 2

Funding: undisclosed amount from
founders; looking for VC funding

Key metric: 20,000 users

URL: www.demandid.com

vote on which acts or performers they want to see, by their own ZIP codes, and to

sell that information to promoters, venues and the artists themselves. “We don’t

want to be a destination site ourselves,” says Cruz. “We just want to put the TourVote

polling application wherever music lovers congregate online.” By aggregating across

sites, Demand ID can gather more (and more valuable) information than it could by

operating at just one destination — much as an ad network benefits by showing ads

and collecting targeting information across multiple websites. The difference is that

TourVote asks permission and gets explicit information, unlike behavioral targeters

who can only watch and guess.

Meanwhile, Cruz hopes, the destination sites will take the information and enrich it

for their own user bases — for example, exposing Juan’s picks to Alice (with permis-

sion), or allowing Alice to see how many of her friends have also voted for kd lang.

Kd lang’s own site could also allow Alice to suggest that her friends vote for kd lang —

but only if they really want to hear her, too. In the end, music promoters and artists

want to know about real demand, not voting by people who won't actually attend a

performance. Of course there’s a soft line here; people who are indifferent to kd lang

could become enthusiastic supporters if encouraged by friends to attend a concert.

The music industry is in a delicate state right now, but Demand ID and similar ven-

tures could help make it into a better industry for its most vital participants: the

artists, who typically make the bulk of their earnings by performing live, and the

audiences. With actual demand being reflected, rather than artificial demand being

created, the “best” artists (as defined by those who excite listeners rather than those

who pay marketers) may get more visibility. Music-industry veterans, while initially

skeptical, won’t turn away good leads, says Cruz. “Nine out of ten promoters, venue

owners and talent agents who start out skeptical become believers upon viewing our

market intelligence prototype,” he claims. “Many of the venues I’ve met with now

have our TourVote polling application on their own websites.”

DECEMBER 2005

RELEASE 1.0

37

38

RELEASE 1.0

He adds, “Today the value chain takes guesses on where shows should be produced.
Then they spend big on advertising to try to reach consumers who will buy tickets.
This is increasingly difficult as consumer attention is fragmented by more and more
media.” And beyond ticket sales, demand to see a particular artist’s live show reflects
demand for ancillary products — everything from CDs and track downloads to
books, ringtones, T-shirts and other paraphernalia.

Right now the service is limited to the Baltimore/Washington area, where Demand
ID is based, and it is working on only 30 websites, mostly for local venues. The com-
pany is currently collecting just a few hundred votes a month as it tries to avoid dis-
appointing people outside its home base. But Cruz has finally decided it’s time to go
for broke and raise serious capital.

We're not sure of the business model. Certainly we'd prefer to see this kind of service
widely implemented across multiple sites. ..and perhaps the only way to get it to be
ubiquitous is indeed to offer it as a back-end service to Eventful, Zvents et al., rather
than as a social tool that is visible to users, but unique to one particular events desti-
nation. Of course, these sites could provide their own competing service for their
users and sell the data not just to musical acts but to all kinds of live performers,
from book readings and magic acts to lectures by Stephen Hawking (who addressed
an audience of 3000 at Oakland’s Paramount Theater recently) or performances by
Tom Peters, Faith Popcorn or other business gurus. We guess that’s why Demand ID
is now trying to move a little faster.

Meetup.com: Bowling together

Meetup’s goal is not to register meetings, but to help create them by providing a plat-
form for local interest groups to form, grow, and arrange regular meetings — or
Meetups, where people can find community, share their concerns, find people of
similar interests and get reinforcement for those interests. It also helps local meeting
organizers band together in larger, topic-specific groups that give their local meet-
ings more weight. That is, one Mom or Mac Fanatic is just a person, but a group of
Moms or Mac Fanatics is a Meetup, and, says founder Scott Heiferman, “Hundreds
of Mom Meetups or Mac Fanatic Meetups, every second Tuesday around the world. .
.that’s a movement.” In essence, Meetup is a way of finding the long tail of diverse
interests locally and making people feel part of the fat front worldwide. (piscLosure:
ESTHER DYSON SITS ON THE BOARD.)

WWW.RELEASE1-0.COM

Heiferman started the company with the mission of using online tools to bring peo-
ple together offline. September 11 had impressed him with the importance of local
community. Then he read Robert Putnam’s “Bowling Alone,” about how Americans
are spending too much time alone. Heiferman wanted to get them together again; all
he needed was a business model. As a former advertising guy (and founder of
iTraffic, the original online ad agency), he came up with a clever idea: Create a plat-
form for people to organize Meetups and sell the “traffic” — i.e. the real people troop-
ing into a Starbuck’s to buy coffee or into Hank’s Bar & Grille to buy beer — to the
venue owners. It was an elegant, clever idea, though in the end it proved too compli-

cated to implement profitably. Among other things, venue owners
weren't eager to host the partisan political meetings that Meetup
was famous for during the Presidential elections.

Nonetheless, even after the elections member numbers continued to
grow, to the extent that last spring Meetup started charging the
community itself for the service. Charging for what used to be for
free took some courage; it cut the number of active Meetups in half.
And six months later, there are now almost as many Meetups as
when it was free, with Meetups and sales growing better than 15
percent each month. Better yet, the people who remain, and the new

MEETUP INFO

Headquarters: New York, NY

Founded: March 2002

Employees: 25

Funding: undisclosed amount from man-
agement, Draper Fisher Jurvetson
and angels including Pierre Omidyar
and Esther Dyson

Key metric: 2.4 million registered users;

over 10,000 paid Meetup Groups
URL: www.meetup.com

members since, are more committed and enthusiastic, says

Heiferman: “Not everything in the world will be funded by ads. We
wanted to be “Of, By, and For The People, so we decided to be funded by the people.
Nothing says “We love your product’ more than paying customers.”

Organizers pay Meetup for the service of helping them find group members and for
the online resources to manage the meetings and create continuity. The organizers
also get a place in a larger community of like-minded Meetuppers for any particular
topic across the planet. To help things along, Meetup both encourages individuals to
sign up for existing groups, and allows them to place alerts, for example: “Know
when Wheel of Time Meetups start near you! See other Wheel of Time Enthusiasts
near New York!”

Time is a key element of Meetups. “It’s like an auction,” says Heiferman. “The dead-
line for an auction, or the constraint that a Meetup is happening at a certain time on
the same day around the world. . that mobilizes people and creates urgency. And it
can make you feel youre part of something bigger.” Now, however, it’s not Meetup
that sets the day anymore; it’s the members de-facto voting: The day on which most
Meetup groups on a particular topic have their meeting has become the “official”

DECEMBER 2005

39

40

RELEASE 1.0

Meetup Day; others are free to deviate, but there is some norm to adhere to. “There’s
this constant tension between freedom and control, “ says Heiferman.

Meetup’s relations with the event portals are friendly, though of course there’s also a
constant tension between competition and collaboration. “We spark events, events
that wouldn't have existed without us,” says Heiferman. “We gladly feed all our basic
information to Eventful and others, and we even pay EVDB for referrals.”

He adds, “At almost any given moment in time, there’s now a Meetup happening
somewhere in the world — a gathering of strangers that share some common inter-
est. Whatever you're interested in, whether you're thinking of buying a Prius or
you ve just been diagnosed with something, there can be a Meetup scheduled about
it. It wort happen just by indexing all the world’s events. There must be a catalyst
like Meetup.com.”

Technorati: Net future value

Tantek Celik came to Technorati after 15-plus years knocking around Silicon Valley
as a developer at Oracle, Apple and Microsoft, and participating in a variety of stan-
dards efforts. He also founded 6prime, a company that developed a tool called Rev
that automatically tracked and managed versions of documents over time by com-
puting and saving changes between versions. It was sold to Aladdin Systems in 1997.

Now Celik has found his dream role at the intersection of many of these interests, as
chief technologist at Technorati. (DISCLOSURE: ESTHER DYSON IS AN ADVISOR.) He’s
designing services to parse formats he’s creating, and giving the blogosphere the
tools to design and analyze itself.

One of the key constructs of Technorati is time. Though blogs are searched for the
key words (and synonyms and related concepts) that a user enters, the results are
ordered by time, by default. Of course, you can explore farther. You can see how top-
ics or particular blogs rise or fall in popularity over time. You can assess the deriva-
tives of growth — not which blog is gaining most readers, but which blog’s percentage
increase in readers over time is highest. You can watch as growth rates increase or
decrease over time. And so forth. Nothing is static in this world.

Celik aggressively values his own time — which makes him willing to spend time now

to develop tools that will save time later. Says he: “If I’m going to be investing my
time to enter or search for event information, I must be able to copy and move that

WWW.RELEASE1-0.COM

information anywhere of my choosing. [Italics his!] Most of these new services, com-
panies, features are all based on the assumption that you'll use their system to solve
all your event-based needs, and that they can provide for all your event-based needs.
But they are wrong.”

He continues: “The software, the services, the features are all irrelevant in compari-
son to the data. That’s why we and apparently other efficient geeks use Word or some
other text editing environment (e.g. Joi Ito uses a wiki) to keep track
of our events. The sides of the road are littered with the remains of

failed (or acquired and neglected) events/scheduling applications, TECHNORATI INFO

services and companies from the ‘80s and ‘90s. Some of them took biesdiainstkenss SanFrancisco CA
all their user’s data to the bit bucket along with them. That pain is Founded: November 2002

still fresh in enough technologists minds, that it will not be forgot- Employees: 30

ten any time soon? Funding: undisclosed amount from

Draper Fisher Jurvetson, Mobius

$ S and angels including Esther Dyson
Thus, he says, the most important aspects of data are longevity,
Key metric: 22 million blogs tracked

integrity and portability. He concedes that these are not new ideas, e a ecni

and there are standards to provide them. However, those standards

have mostly lacked an important component — users. By and large,
time and contact standards — the infamous .vcf card — just haven’t gotten broad trac-
tion. Instead, lots of proprietary systems — most notably Outlook — hold sway.

Enter microformats. Unlike most standards, they are simple, designed for humans,
and relatively easy both to create and to use on the Web.

One of the first microformats was hCalendar, conceived in 2004 at O’Reilly’s Foo
Camp. It’s basically a short, simple set of metadata specs for describing an event,
including title/summary, start/end times, location, description and other common
fields, all based on the IETF RFC 2445 “iCalendar” standard.

Using the open-source hCalendar-creator available at microformats.org, any blogger
can post and discuss events on her blog in a form that can be easily discovered,
parsed, aggregated or subscribed to by other services or by other users using a sub-
scription proxy available free from Technorati. And any application developer can
use the format in order to make its event/schedule data easily tradable with the end-
user’s other applications. Eventful, Zvents, Upcoming.org and Laughingsquid.com
all support hCalendar.

DECEMBER 2005 RELEASE 1.0

42

RELEASE 1.0

Indexing and searching these event sources across the Web is next. Says Celik:
“Whatever event/calendaring applications/service/feature you are using, the most
important aspect of it is how it treats the data you have entered — and thus the time
you spent. And the most important aspect of your data itself is how easily you can
reliably move it around. We’re just beginning to see the impact.”

Adding Value With Time

The first part of this issue covered applications that concern time in the traditional
sense — calendars, scheduling, events. This second half looks at applications that use
time more as a mechanism to do other things (although the distinction can some-
times be blurry). Specifically, the first section focuses more on the absolute place-
ment of things in time: when they happen, which things happen together and how
to manage the activities of people through time. Call it time as location.

In this second part, we look at applications that use time more as a measure, look-
ing at the frequency of events, the rhythm of activity and patterns of use. Call it
time as distance.

Worktopia, though it is concerned with scheduling of space now, fundamentally
looks at time and space as commodities in a spot market. It will ultimately be con-
cerned with pricing and yield management, watching the rise and fall in demand for
office space in real time.

Detection and recognition

So, how can time help us manage other things? In large part, it’s an important ele-
ment in defining and understanding what’s going on in the real world. The better we
can identify patterns and objects in the increasing amounts of data we collect and
use, the easier it is for us to find things, predict things. . .and to let software handle
those things automatically.

Rather than have humans looking through information, computers can do so, and
simply deliver the high-level information: “These pictures all depict your uncle
Sanjay.” Or, “These few hundred Web-surfers are probably interested in buying a car
[because they have visited several car sites in the last three days].” Or, “There’s an
unusual spike in purchases of fever medications in New Jersey; perhaps there’s some

WWW.RELEASE1-0.COM

infectious disease going around.” Or, “Someone seems to be trying to get into our
database of Christmas merchandise orders.” And if you can define things precisely
and explicitly enough, software can deal with many of the details, from tagging pho-
tos to fending off computer-virus attacks.

Indeed, as software starts to handle the real world, it needs to deal with unprecedent-
ed volumes of data, whether it’s the myriads of data points generated by RFID, the
hundreds of e-mails generated by humans or the millions of clickstreams generated
by all of us on the Web. Or the millions of people and resources and the events in
which they intersect — meetings, e-mails and moments captured in photos.

Running through this whole issue are applications of pattern recognition, detection,
figuring out intentions, etc. That’s because the underlying task is to figure out what’s
going on and to represent it. In order to make sense of what’s happening in the real
world and to deal with the proliferation of events at scale, we need automated tools
for recognizing and annotating them. That’s everything from detecting references to
meetings in e-mail (Zimbra), figuring out what changes someone meant with a doc-
ument edit (Curtis) and detecting people in photos (Riya, below), to recognizing
objects in videos (EventWeb) and of course seeing patterns in blogs (Technorati),
disease incidence (CDC, below), security anomalies (Sana Security, below) and buy-
ing intentions (Revenue Science and Dan Doman, below).

Worktopia: Office arbitrage

Our increasingly hurried society measures everything in smaller and smaller chunks
of time. Two centuries ago, citizens were happy to listen to three-hour speeches from
politicians; now they tune out after a two-minute soundbite. Moreover, in the old
days, we use to think of things as things. You could own them or trade them, but only
economists and landlords thought much about the value of most things over time.

In the “80s and ‘90s we discovered yield management (SEE RELEASE 1.0, FEBRUARY 1989).
Now, we're rediscovering spot markets, not just for commodities and advertising
minutes, but for vacation rentals, temporary offices and jet planes. Soon to come, we
believe: A spot market for speakers — matching an expert on a layover at DFW with a
Dallas user group, for example, or a recent author of a book about witches with a
series of Wiccan Meetups. The promise of car-sharing continues to be just a
promise, but ride-sharing services may gain more traction as gas prices go up and
express lanes become more appealing.

DECEMBER 2005

43

Meanwhile, there’s Worktopia, bringing efficiency and market timing to the world of
temporary office space.

Worktopia founder John Arenas previously started another office-space company,
called Stratis Business Centers, back in 1997. In the ‘80s he had started off working
for Turner Construction and then for Citigroup’s workout loan group, so he had a
good understanding of real estate, and specifically of office space, from a practical
and a financial/valuation point of view. His initial idea was to create a network of
office space facilities that would license short-term space (hourly, monthly and
annually) to customers nationwide as an alternative to conventional long-term lease
commitments. In order to grow without much capital, he created a

WORKTOPIA INFO

franchise network. And in order to offer something special, Stratis
developed the technology to control the office space and equipment

Headquarters: Hanrison, Ny centrally, including interfaces with Nortel phone systems, Pitney

Founded: October 2004

Employees: 7

Funding: $1.25 million in convertible debt

Key metric: 10,000 completed searches

Bowes meters and Konica copiers as well as its own proprietary
reservation system.

ane St militon transäctiornrevénüë The firm grew nicely, to 20 locations in 30 months, but Arenas still

per month

URL: www.worktopia.com

felt hampered by lack of expansion capital. In 2001 he sold out to

UK-based Regus, the world’s largest temporary office-space compa-

44 RELEASE 1.0

ny, then operating in 50 countries and which wanted to expand
more quickly in the US. But, he recalls ruefully, “I wanted to make it
tick; they wanted to make it sizzle.” In short, Regus overexpanded in the US and
Arenas spent most of his time restructuring leases rather than building the network
of on-demand workspaces he dreamed of. In the end, he had the “learning experi-
ence” of leading Regus through bankruptcy in the US in 2003.

By January of 2004 he had brought Regus out of bankruptcy and, his job done, he
left in February to start Worktopia. This time, he’s operating as a marketplace rather
than as a workspace supplier. The properties on offer with Worktopia are more var-
ied than at Stratis or Regus: from office business centers offering short-term office
space, to last-minute hotel meeting rooms, movie production and staging facilities
on demand... and of course university facilities outside class hours (such as
Stanford’s Schwab Residential Center, where we are holding our When 2.0 workshop
— but not — yet — a Worktopia customer).

At this point, however, short-term office space and hotel meeting rooms are the

biggest part of Worktopia’s inventory, in part because hotels and business center
operators understand the short-term and last-minute resource models better than

WWW.RELEASE1-0.COM

most conventional real-estate owners. But that also means they are the most efficient
part of the market; over the long run, as Worktopia introduces other players into the
game, the opportunities for arbitrage between unused space and needy customers
should be even greater.

An MIT/Gartner study estimates that about up to 10 percent of office space will be
used on a flexible/short-term basis by 2010, up from only about 1 percent today.
There will be some growth in demand, but this means that a significant proportion
of the nation’s 3 billion square feet of Class A office space market is currently under-
used, misallocated or overpriced. Or looked at another way, the current market for
short-term office space is about $2 billion, and it’s growing rapidly. The day-meeting
market is already $7 to $10 billion, according to whom you ask.

Space to match pace

In essence, just as eBay allowed people to turn stuff collecting dust in attics into
valuable inventory (SEE RELEASE 1.0, APRIL 2004), Worktopia is allowing not just
unused office and meeting space, but the chunks of time that any particular piece of
space is unused, to be turned into valuable inventory. Industries such as airlines —
notably! — and hotels, where fixed costs were higher, got into this early on; they
couldn't afford to hold airplane seats or rooms and not use them. But most other
companies treat real estate as an unoptimized asset, and only recently started think-
ing about it strategically.

And now, says Arenas, “Space users want a sense of control. They want real-time
information to compare alternatives, from a trusted, unbiased source. Typically,
they’re buying with a lead time of weeks — not two years like for a convention — with
no time for research.” Until now, there was no marketplace where they could find
this. Many of them resort to calling hotels directly — hardly an unbiased source — or
rely on local colleagues.

Enter Worktopia — which opened to the public last June. So far it is handling about
12,000 searches — or $1 million in transactions — per month. Its revenues include 8-
to-10 percent commissions on temporary office space (payable by the supplier).
Hotels and other meeting-room marketers can also subscribe for $2000 to $4500 per
hotel per year to offer rooms commission-free (which overcomes the touchy issue
that hotels make much of their money not on the meeting rooms but on the ancil-
lary services and accommodations).

DECEMBER 2005 RELEASE 1.0

45

46

RELEASE 1.0

Finding the long tail of customers is relatively easy compared with finding the long
tail of suppliers — and certifying quality. Those who rent out the space are grateful to
find Worktopia. But even among big suppliers such as hotel chains, only Hyatt and
Hilton have central databases of availability. For now, Worktopia is building out a
supplier base through direct sales. Eventually it will probably use customer feedback
to create a reputation system, following the lead of many other marketplaces.

As for sophisticated yield management, that’s an obvious development but still in
the future. Although a couple of the hotels have sophisticated pricing algorithms and
booking-limit strategies in place, most suppliers and customers still work by feel in a
fragmented, opaque marketplace. Eventually, Worktopia hopes to provide its cus-
tomers (on both sides) with analytical tools to help them operate more effectively in
a more transparent market. But right now, both sides are losing as much space goes
unused, while the space that is used is often incorrectly priced. A better understand-
ing of both time and space would help a lot!

Centers for Disease Control and Prevention: Time is meney life

The Centers for Disease Control and Prevention is an old, established institution,
founded in 1946 — though young in government years. Its initial mission was to con-
trol malaria; its initial staff of 400 comprised mostly entomologists and engineers,
and only seven doctors. Since then, it has grown to 9300 people, including many IT
specialists serving people such as Robert Pinner, MD, director of the Office of
Surveillance in CDC’s National Center for Infectious Diseases.

Pinner himself trained in internal medicine and then became a specialist in infec-
tious diseases. For him, medicine is an intellectual pursuit as well as a service career.
As a kid, he says, “I did all the things you can do with baseball cards and batting aver-
ages. And now, I like math more than most doctors. I’m fascinated by the challenges
of understanding how diseases spread, how they can be controlled, and evaluating
whether control measures are working. For example, in evaluating the efficacy and
impact of vaccines, the math is the least of it. The harder challenge is design of stud-
ies in the real world that can give answers that will be useful for public health. In
particular, post-licensure evaluation of drugs and vaccines is a tough challenge.”

After getting his MD from Emory, he stayed in Atlanta for medical residency, a couple
of years as a primary care physician and postgraduate clinical training in infectious
diseases. At the time, AIDS was just making the transition from being viewed as a
syndrome to being understood as a disease caused by a particular virus. As an infec-

WWW.RELEASE1-0.COM

tious disease, notes Pinner, it was something of an anomaly: “Formerly, when faced
with a problem in several organ systems, it made the most sense to think about how
to get to a unifying diagnosis. But once a patient was diagnosed with AIDS (usually
because of an opportunistic infection), you had to consider not only HIV infection
itself but the variety of opportunistic infections that patient was susceptible to.”

Living in Atlanta, home of the CDC, Pinner gravitated towards the CDC’s Epidemic
Intelligence Service, a two-year applied epidemiology training program (sometimes
called “shoe-leather epidemiology”) which he took from 1988 to 1990.

His work now involves surveillance, managing CDC’s Emerging Infections
Programs (EIPs), a network of CDC and 10 state health departments and their acad-
emic collaborators, with a $20-million annual budget. One basic tool is data and
charts depicting changes over time in everything from reported diagnoses, to pur-
chases of drugs, use of specific kinds of lab tests and a variety of other indicators of
underlying diseases and symptoms.

There are also lots of practical questions about getting the data: How do you detect
disease in the first place? In the balkanized US health care system (SEE RELEASE 1.0,
JANUARY AND SEPTEMBER 2005), there’s no single database of health records, let alone
common standards for reporting disease.

Desperately seeking symptoms

But there are ways to pick up hints, he points out, other than a central database. For
example, he was part of a team responsible for monitoring cases in New Jersey dur-
ing the post-9/11 anthrax bioterrorism-related outbreak, when six cases of anthrax
occurred resulting from circulation of contaminated mail. The team investigated the
outbreak, conducted environmental sampling of postal facilities, and did surveil-
lance to monitor for additional cases. ” This is an example of where electronic
exchange of clinical information from emergency rooms would have saved a lot of
labor-intensive surveillance work if it had been available,” says Pinner.

And sometimes there are false positives, such as a seeming outbreak of heart prob-
lems, as signaled by a three-day surge in the number emergency room complaints of
chest pain in the late summer of 2004. The cause? As far as the New York City health
department could figure it out, it was former President Clinton’s announcement that
he would undergo heart surgery, which evidently either sensitized people to their
pain or triggered them to take it more seriously. Often, it may be easier to identify
patterns than to figure out what caused them.

DECEMBER 2005 RELEASE 1.0

47

48

RELEASE 1.0

Timing matters, too. The earlier you detect something, the better off you are, in the-
ory. But in practice, it’s more difficult to be sure what it is. For example, data from
over-the-counter pharmacy sales for drugs to treat flu symptoms may be available
earlier than specific diagnoses of flu cases, but they are not at all specific for influen-
za. And early on, it’s hard to know whether an increase in some indicator is an
anomaly, or the start of a steady rise in, say, flu cases.

So, says Pinner, he and epidemiologists in general continually face a balancing act
between time and uncertainty. “The better we get at early detection, the more ulti-
mately harmless ’threats’ we will detect.” For example, you might see a “blip” because
a data entry person was catching up with chores after being out for a few days, or
because a process that had been down, went back up and was catching up; or if you
were monitoring over-the-counter drug sales, you might pick up a “threat” that was
nothing more than a sale at CVS. False alarms can cost money and disrupt people’s
lives; missed alarms can also cost money and result in needless death. Concludes
Pinner, “Lots of experimenting is going on now with efforts to identify earlier signals.
People need to evaluate carefully what works and what doesn’t, what’s effective — and
cost-effective — to detect and what isn’t.”

Sana Security: Time and security

Steven Hofmeyr, a computer scientist specializing in immunology with ties both to
the MIT AI Lab and to the Santa Fe Institute, founded Company51 in 2000 to apply
some of his insights about biological immune systems to computer security (SEE
RELEASE 1.0, MARCH 2002). The company, now called Sana (as in “health”) Security, has
just released a standalone desktop product called SafeConnect. (DISCLOSURE: ESTHER
DYSON IS AN INVESTOR.) Hofmeyr is now a consultant to the company and is working
on a new start-up, exploring new ways of building distributed storage systems.

Underlying Sana’s products is Hofmeyr’s understanding of the “organic” nature of
security. Time is a key factor in assessing behavior as suspicious. Someone sending
out files at 4 pm is probably normal; the same person sending the same files at 4 am —
assuming he’s not traveling in some other time zone — is suspicious. Any security sys-
tem probably assesses such factors and then — if it’s not in the system — some human
will check to see where Juan was that particular morning. Was he traveling in France
on business? Or was he — supposed to be — on vacation? It’s the combination of fac-
tors that counts. But there are many more subtle uses of time, says Hofmeyr.

WWW.RELEASE1-0.COM

Any suspicious event becomes more suspicious if it occurs along with another suspi-
cious or non-routine event — for example, a file upload that happens around the
time of a door-lock malfunction, or the attempted use of an expired password short-
ly after an employee retires.

But, notes Hofmeyr, “After the first detection of something odd, you

can’t act immediately or you'll end up on constant alert. You have to
. po ; . H SANA SECURITY INFO
wait to see if it’s a real threat. . .but you can’t wait too long...

Headquarters: San Mateo, CA

: n : Founded: October 2000
He resorts to immune-system analogies. “The immune system

x R r . Employees: 60
exploits time. You bet that you have the time to build up an effective

response. . . Besides, if you react immediately to something you

Funding: undisclosed amount from El

Dorado Ventures and angels includ-

think is bad, that reaction itself could hurt you, so you have to wait ing Esther Dyson

for some damage before reacting. The problem is that what appears Key metric: more than 200 customers in
unusual may be benign — so the immune system waits until damage 10,009 0cations

occurs because then it can be sure that the unusual behavior is not URL: www.sanasecuritycom

benign — but then of course the immune system has to play catch-up
— it’s a case of giving the pathogen ‘enough rope to hang itself”
Knowing precisely what that timing should be is what distinguishes good security
from ineffective responses.

Now that Hofmeyr is leaving, one of Sana’s key employees is Matt Williamson, who
came up with the concept of “virus throttling” at HP Labs. Viruses are most harmful
(obviously) when they spread rapidly. . .and that’s something that “normal” pro-
grams don’t do. Even P2P music files, spread by individuals, don’t spread as fast as
viruses; most P2P software limits the number of concurrent downloads from one
computer to a just a couple.

An individual using a computer, even a busy browser, is unlikely to connect to more
than five or so new addresses in a minute. (Any security system knows how to make
an exception for a mass mailer, though an ISP’s security system monitoring a cus-
tomer base of consumer machines might rightly not make such an exception.) “For a
virus, that’s slow. An infected machine might try to connect to thousands of other
systems in a minute,” says Hofmyer. So you can just default to prevent such behavior
by limiting the number of new connections a computer makes per minute, and alert
a monitor when it is attempted.

DECEMBER 2005 RELEASE 1.0

49

50

RELEASE 1.0

“That slows the propagation of viruses, but doesn’t bother people. We exploit the
different meaning that time has for people and machines,” he says.

Don't kill the cop

Yet the similarities are useful too. There’s a truism in immunology that pathogens
don’t want to be too harmful because they want their hosts to survive. As software
becomes more and more malicious, it also becomes benign in some way because it
wants the host to stay alive. So a virus inside a host population — if the population
doesn’t fight it — tends to become benign. But if other pathogens are around, then it
may become virulent — if only because it has little to lose. “Whoever kills the host
first wins, because he gets the most out of it,” says Hofmeyr. And then there’s the
phenomenon of pathogens attacking one another, while the host is simply an inno-
cent bystander/environment. All this happens over time; the security expert’s task is
to figure out these cycles and exploit the vulnerabilities.

In the spyware world, for example, there is the phenomenon of spyware packages
trying to de-install one another in order to gain exclusive access to the host.

Or, notes Hofmeyr, “there were 11 variants of the Zotob worm competing ferocious-
ly, all trying to delete each other. They were so busy harming one another they may
have done less harm to the host — even though they may also have spread faster in an
attempt to get to the vulnerable hosts first. Imagine a worm that spreads and makes
you less vulnerable to other guy [worm], but then it does its own damage three
months later.”

And finally, he notes, time is important in consumer anti-malware. Most security
software, he points out, lets you know after the fact that you have a problem.

By operating in real time, Sana’s tools have two benefits, aside from the obvious one
of averting damage quickly: The user is more likely to be able to determine the
source of the malware: “What website are you surfing?” as opposed to “What web-
sites have you visited since the last time you're sure your machine was clean?” And
also, because the tool is watching behavior rather than results, it has more under-
standing of what the malware did so that it can reverse the effects. “Because we mon-
itor the system as it’s running, we can see what it did, take off the bad stuff and clean
up the mess. It’s a pleasure to watch this in operation cleaning off stuff. For example,
it knows which registry keys a program changed and how they were changed so that
it can reverse the changes. That could be impossible to figure out after the fact. “

WWW.RELEASE1-0.COM

Riya: Eyes only for you

Munjal Shah started Ojos — now renamed Riya, after the daughter of a co-founder —
after co-founding and running Andale (the company that manages the auction
process for eBay sellers) from 1999 to 2004, bringing it to profitability on double-
digit-million revenues.

The idea came before the technology: Newly retired, Shah had been spending time at
various photo sites and poring through his own collection of 30,000-plus photos,
and was discovering how tiresome it can be to tag all those photos. He started with a
vague idea of automating the tagging process, and he had the time and money to do
some research. Among the interesting findings: on the typical consumer photo-shar-
ing site, 80 percent of the photos contained faces of people, as opposed to distant fig-
ures or no people at all. (That percentage is lower for professional sites or for Flickr,
incidentally.) Shah also surveyed consumers in Korea, which he considers a proxy for
the cell-phone culture of the future. He found numerous people who had more than
10,000 photos, mostly of their friends in a variety of situations.

He decided that figuring out how to identify all those people in the photos and tag-
ging them semi-automatically would be a big win. The next step was to find the
leading facial recognition experts; he found Burak Gokturk, an

experienced, widely published expert in the field. Development
. ar l RIYA INFO
started in October 2004; the service is now in alpha. (piscLosure:

Headquarters: Redwood City, CA
Founded: March 2005

ESTHER DYSON IS AN ADVISOR.)

$ Dut F 3 Employees: 22
Clearly, the thing that most distinguished the photos was time — easy y

Funding: $4 million from Blue Run

to get — and the people in them — hard to get. Thus, the system uses

Ventures and Leapfrog Ventures

timing and other cues to improve the accuracy of facial recognition. Key metric: not yet launched
It also leverages the existence of social networks by getting individu- URL: www.riya.com

als to share their “training sets” — that is, a set of photos that can
train its engine to recognize that particular person. Those sets are
sent to Riya, which then uses, say, Alice’s photos in order to recognize Alice in the
photos from other Riya users she designates as friends. That is, her friend Juan gets
to use Alice’s training set (via the Riya service) on his own photos. It makes sense:
Friends take photos of friends, so the network supplies both the data and the con-
straints to support effective recognition.

Time, of course, figures in things too: Photos taken close together are more likely to

contain the same person than photos taken apart. The system applies any tags it gen-
erates across similar photos taken at the same time. It also uses time to weight train-

DECEMBER 2005 RELEASE 1.0 51

52

RELEASE 1.0

ing-set examples, so that, for example, if you grew a beard in October 2003, then
your examples from that time will be weighted higher in assessing photos from that
period onward. Now add to that (someday!) a link to everyone’s calendars, and the
recognition accuracy will become even higher.

Over time, even as Riya will use social networks information to enhance its recogni-
tion capabilities, other tools may work in the opposite direction. Whom do you take
the most pictures of? Who was seen next to whom most frequently? (Will Riya evi-
dence end up in divorce court someday?)

And of course, yes, there are endless privacy issues floating around, starting out with
who owns the training set. Basically, it’s Riya plus the individual concerned — Alice,
for example.

Riya does have a site of its own that retains the training set for each recognizable
user as well as some user photos, but we expect it to succeed primarily on the basis
of a back-end distributed model that supports a broad range of third-party photo-
sharing sites.

Here’s how it will work: Alice gives Riya access to all her photos on, say, Flickr and
Picasa. Following *their* rules for acknowledging friends (which may need to be
updated to accommodate Riya), it recognizes all of Alice’s friends who have agreed
to let her use their training sets on Riya. Then it uses each service’s API to apply the
Alice tag automatically to all his photos that contain Alice.

Of course, Riya will control a hugely valuable (and vulnerable) asset in the ability to
recognize all those people. And by collaborating with third-party sites rather than
competing with them, it has some hope of becoming a central utility — with all the
responsibilities and accountability that that implies.

Time and Intentions: The Commercial Meaning of Time

Nothing interests marketers more than customers’ intentions. Advertisers and pub-
lishers spend millions trying to divine them. But those intentions change over time.
What is the curve in time of a person’s intention to buy, say, a car? You're pretty sure
Juan wants to buy a car, but if you try to sell him one too soon, it’s like trying to rush
a first date through dinner. You have to wait for the kiss until dinner is over. . .but if

WWW.RELEASE1-0.COM

you wait too long, someone else will already have stolen your date or sold Juan a car
before you can make your offer. Figuring out those curves is not something you can
do in a vacuum or even in a psychologist’s study: You just have to watch how (lots
of) people behave.

Learning how to time things better will be a key challenge for the growing field of
behavioral targeting — something of an extension of the Amazon recommendation
engine on one side and cookies and spyware on the other — which is just beginning
to emerge as a respectable market. Revenue Science and Tacoda are doing it with
publishers; Compete Inc. is doing it in conjunction with UPromise and, it hopes,
other marketing partners. ChoiceStream is doing the equivalent of behavioral tar-
geting for commerce (for customers such as Yahoo! and AOL), and others are apply-
ing it to search.

So far, the targeting has mostly to do with detecting the presence of potential cus-
tomers and showing them ads without much respect for timing. That’s partly
because the volume of good data is so low (or almost drowned out in noise) and the
practice is so new.

There are also a broad range of shopping tools designed to show shoppers competi-
tive offers (SEE RELEASE 1.0, JUNE 2005), but most operate in a fairly rote fashion, dis-
playing a range of competing offers from various vendors and merchants
simultaneously. We haven’t yet seen the user-controlled tool that can figure out what
you want before you know it yourself, that can guess your proclivities rather than see
your explicit intentions. You may or may not want to discover your own hidden
greedy child, but marketers do.

Thus you can expect to see a lot more behavioral targeting, even though in the past it
has been associated with adware/spyware companies. You don’t need a download to
recognize a user’s cookie as he moves from site to site, and, as consultant Dan
Doman (below) points out, “You don’t need to know any personally identifiable
information at all. You just need to know where you saw the person before — but not
who the person is.”

Yes, behavioral targeting has bad associations, but it’s one of those wonderful things
that makes just about everyone happy: more relevant ads for the consumer, higher
revenue per ad for the publisher, and higher click-through rates on the ads — which
generates an increased ROI for the advertisers even at those higher prices.

DECEMBER 2005 RELEASE 1.0

53

Revenue Science: Take me to your gold mine!

Revenue Science began life as Digimine, a data-mining firm that analyzed user
behavior on behalf of e-commerce sites. The company recognized the big win it
could gain if it could use behavior data to generate incremental revenue for media
sites, and in 2003 it changed its name and its focus to work with publishers and
advertisers rather than just individual commerce sites. Revenue Science media cus-
tomers now include Fortune 500 advertisers and media properties such as The Wall
Street Journal Online, ESPN.com and Jumpstart Automotive Media.

Once Revenue Science had a following of premium properties using its behavioral
segmentation within their sites, it embarked on a plan to persuade its publisher-cus-
tomers to let it use their data on behalf of other customers across a network. That is,
if Juan visits the auto pages of a leading auto research site, Revenue Science can use

that information to show Juan car ads not just on other pages within the auto site,
but also on the weather pages of, say, the Washington Post. Two important points:

REVENUE SCIENCE INFO

Headquarters: New York, NY and
Bellevue, WA

Founded: March 2000

Employees: 78

Funding: undisclosed amount from
Mayfield, Mohr Davidow, Integral
Capital Partners and Meritech
Capital Partners

Key metric: access to more than 1000
websites that reach more than 100
million users worldwide

URL: www.revenuescience.com

One, the auto publisher doesn’t share its data, but it does allow
Revenue Science to use conclusions based on the data — that Juan
may want to buy a car — to serve an ad for another publisher. Two,
the auto site gets to share in Revenue Science’s revenue for that ad.

Over time, that proposition has proved compelling for several cus-
tomers, and more are likely to sign up. Revenue Science has just
launched a network with dozens of publishers providing data and
many more agreeing to serve ads. The audience reach for the net-
work is already greater than 50 percent of the US online population.

The economics are compelling. A regular ROS (run of site, or gener-
ic) ad on, say, the news page may go for several dollars per thousand
impressions. An ad targeted at a potential car buyer may bring in 10
times that (especially since most “inventory” on car sites is not just

expensive but often unavailable). The net result is that the auto site gets money it

wouldn't have earned otherwise and loses nothing directly.

For example, says Revenue Science’s Omar Tawakol, senior VP and general manager
of Behavioral Targeting Marketplace, “For most response advertising, you can get
pretty much double or triple the response of not using targeting at all, without even
paying much attention to timing details.” Yet, he expects, timing details will matter. If
you can get three times the detail without effort, imagine what you could do if you

54 RELEASE 1.0

WWW.RELEASE1-0.COM

could catch buyers at precisely the right time. The limit, of course, is the actual num-
ber of buyers waiting to be discovered.

Because it’s a new field, with most such campaigns just rolling out, there’s still a lot
to learn. Revenue Science is currently serving several billion behaviorally targeted ad
impressions a month across a couple of hundred publishers. That generates a lot of
data to analyze. Which combination of factors yielded click-throughs, and which
didn’t? “Psychology is driving user behavior, but it’s the math that discovers it,” says
Tawakol. “However, learning something conclusive is hard because there are so many
variables. It’s almost depressing if you're a scientist. The noise is so tremendous that
you need to be very deliberate about your experiments. There can be a factor you're
not even aware of — such as the amount of white space in two different ads — that
affects your response rate more than the variable you're measuring.”

The rhythm of buying

Right now, Revenue Science (and its competitors) are just experimenting. “We get
such an uplift doing simple targeting,” says Tawakol, “that we haven’t yet refined the
timing.” However, they are working on it. “Recent is usually better, but you don’t
want to miss people by making the window too short.” Of course, the precise figures
will end up being proprietary, though the overall approach is clear: “We’ve been
experimenting with the time window for cars. Six months is too long; two weeks
may be too little. Six weeks is probably about right. Certainly, it’s not linear; there’s a
decay function which means there’s a rapid drop-off after a certain point; we just
don’t know where it is, yet.”

Moreover, the timing will be different for different items. For now, direct response
marketers are focusing on big-ticket items such as cars, mortgages, insurance and
travel. These bring the greatest return on advertising investment, and they are also
considered purchases. People don’t usually decide to buy a car in the morning and
arrange delivery that night.

Time also plays another role in advertising. The trick isn’t just to figure out which
customers have which intentions; it’s the very practical question of optimizing the
use of ad inventory. Suppose Juan shows up at your site. Based on his recent behav-
ior, he could be in the market for a car. Meanwhile, there’s Alice. Based on her behav-
ior, she definitely wants a car and may want a mortgage. Should you show Alice the
lower-yielding, but surer-bet car ad and “waste” the mortgage ad on Juan? Or should
you show Juan the car ad and Alice the mortgage ad? It’s not a question of principle;

DECEMBER 2005 RELEASE 1.0

55

56

RELEASE 1.0

it’s a question of statistics — and the better your data, the better you can assess the
potential of each strategy.

Tawakol himself is a reflection of how online advertising has evolved over time. Pre-
viously, he was co-founder and CEO of CoRelation, a behavioral segmentation and
personalization company acquired by Revenue Science in 2001. Before that he man-
aged consumer messaging products at Netscape spinoff Liberate, which provided
interactive TV technology for AOL and Comcast, among others. He began his career
as an AI researcher at Stanford University’s Logic Group and HP Software Labs.

Dan Doman: Intentions and inclinations

As we said above, “The limit, of course, is the actual number of buyers waiting to be
discovered.” But there’s another approach, which is to create buyers by divining their
proclivities and making them offers that will induce them to buy things they weren’t
actually looking for. (Yes, and you thought this is what all advertising was about!)

Dan Doman, a happily independent consultant who formerly worked for
DoubleClick and Direct-Revenue as well as Mediaport (a dot-com ad-agency joint
venture that imploded after 9/11), has spent much of the last decade working on
advertising engines and exploring the differences between intentions and proclivi-
ties. An intention is an explicit plan to buy something, as when you are looking for a
car or a good deal on a Nano. A “proclivity” is a tendency (or a vulnerability) that
can be triggered by a good ad or even just an offer. “For example,” says Doman, “I
walked past Best Buy five times in the last week with no intent to buy a Nano [which
they were showing in the window]. But I did have a proclivity — being a music-lover
and a boomer male — and finally I went in and bought one.”

You can guess individuals’ proclivities from their demographics, he points out:
Teenagers like ring tones; women like perfume; young single men like violent games
and DVDs. And you can guess their demographics and proclivities — as well as their
intentions — by the sites they visit. There are also negative cues for marketers, such
as someone Visiting a high-risk loan site (unless youre in the high-risk loan busi-
ness yourself).

Of course, all these are generalizations, but marketers work with statistics, not with
unique, discrete individuals. And from a privacy point of view, that’s a good thing.
You don’t need to know who they are, says Doman. “You get a certain liberation
from identity when you are treated as a statistic.”

WWW.RELEASE1-0.COM

But proclivities and especially intentions change over time. Figuring those things out
is the marketer’s black art. “There are different wavelengths to intentions and pro-
clivities, varying by person and by product,” says Doman. It’s fairly simple to show a
car ad to someone who was browsing Volkswagen’s, but when is the best time to do
so (if you have enough traffic and enough visitors to pick)? “It’s like seduction,” says
Doman. “You can’t start right away with the offer.” From an advertiser’s point of
view, it’s best to let other advertisers or sites educate the consumer, and then show
up with your ad at the time the prospect is ready to make a decision. When is that?

Says Doman, “You can work a lot of these things out. The more data you have, the
better you can fine-tune your timing. You can also come up with things that aren’t
obvious.” For example, he says, someone looking at DVDs is probably looking to buy
a movie. But someone looking at flat-screen TVs is not likely to be looking for a
small-screen DVD player, whereas a frequent traveler who is browsing DVDs may
well have a proclivity for a compact DVD player. However, Doman notes, “The accu-
racy of proclivity measures also decays over time. If you don’t allow for proclivity
decay, the real peaks get lost in the noise. This is also why negative indicators are so
important. The guy shopping for high-end ‘guy-oriented’ consumer electronics such
as large-screen TVs suddenly shifts into a different demographic when he starts
looking at baby furniture, carriages and family/new-born stuff. But demographics
persists: These same BOBOs [Bohemian Bourgeois, from David Brooks’ book
“BOBOs in Paradise” ] will translate their need for Bang & Olufsen stereos, granite
kitchen counters and showers with skylights into high-end Hummer baby carriages
selling for $800 to $1000. The same lust [and income] is there, but the proclivity
manifests itself in different objects.”

However, notes Doman, the long-term advantages will accrue to services that ana-
lyze their data to divine patterns that are not immediately visible or guessable — the
online equivalent of the fabled insight that men buying diapers late at night have a
proclivity to purchase beer as well. . leading thousands of groceries to stock the beer
with the Pampers.

“The point of statistically based behavioral targeting is that it helps you to refine
your targeting using combinations of patterns that are not so obvious anecdotally.
This is where computers and data mining comes in. The traditional marketer knows
these things at a high level through intuition or experience, but doesn’t have the data
to make the right trade-offs. Grinding through vast amounts of data can reveal pat-
terns that are less obvious. If you can figure out exactly when they peak you can get a
better return than someone working just with averages.”

DECEMBER 2005 RELEASE 1.0

57

Conclusion: The End of the Beginning

Everything you have read so far simply sets the scene — both for our When 2.0 work-
shop and for continuing developments in the role of time in software. In the end the
challenge isn’t managing time, but managing other things within the constraints of
time. Beyond the calendar, scheduling and events tools described in the first half,
there is a tremendous amount of variety in how time can be used, as illustrated in
the second half. As we move into a world of cell phones, GPS devices and RFID more
and more of our systems are collecting continuing streams of data in real time.
Without a better understanding of how things change through time, we’ll become
overwhelmed with waterfalls of meaningless data. But with better historical data and
real-time analysis, we'll be able to understand our world much better, seeing not
only what is happening but what’s about to happen and what has already happened.
Just as you can get value from understanding that a text string is in fact a crisp and
parsable description of an event with a defined location, participants and timing.

The ability to see more things more clearly through the lens of time is invaluable. It
gives us better information with which to make decisions, and it will allow us to
automate more decisions. If we can define situations and rules clearly enough, we
can have software execute them.

In short, time is the measure of anything else we do or have. Any child instinctively
knows that a bicycle today is worth more than a bicycle at Christmas. And any par-
ent knows through experience that back-to-school is the time to buy clothes, that
the teen years are troublesome, and that childhood friendships (like adult ones) wax
and wane over time. For some people, it takes five minutes in the morning to make
six useful decisions, while five minutes at night are good for little more than watch-
ing Jon Stewart or Adult Swim or slitting open paper envelopes. For other people,
vice versa.

People also know how challenging it can be to schedule all those magic moments;

that’s why rich or important people have appointments secretaries. Their money is
fungible; you can trade it freely. But their time is not. $r 1.0

58 RELEASE 1.0 WWW.RELEASE1-0.COM

Resources & Contact Information

Dan Doman, Consultant, 1 (212) 289-1024; dan@danieldoman.com

Brian Dougherty, CEO, Airena, 1 (925) 677-0600; fax, 1 (925) 677-7360; brian.dougherty@airenainc.com

Robert Pinner, Director, Office of Surveillance in the National Center for Infectious Diseases, Centers for Disease
Control and Prevention, 1 (404) 371-5360; fax, 1(404) 371-5445; rpinner@cdc.gov

Dean Ratcliffe, CEO, Columba Systems, 44 (20) 7739-9902; dean.ratcliffe@columbasystems.com

Ben Cruz, Founder & CEO, Demand ID Systems, 1 (410) 466-7467; fax, 1 (301) 330-5962, bcruz@demandid.com

Brian Dear, Founder & CEO, EVDB, 1 (858) 964-0697; fax, 1 (858) 964-4640; brian@evdb.com

Thomas Moran, Distinguished Engineer, IBM Almaden Research Center, 1 (408) 927-3844; tpmoran@us.ibm.com

Scott Heiferman, CEO, Meetup, 1 (212) 255-7327; fax, 1 (212) 255-7310; scott@meetup.com

Pavel Curtis, Software Architect, Microsoft, 1 (425) 722-0920; fax, 1 (425) 936-7329; pavelc@microsoft.com

Ray Ozzie, CTO, Microsoft, 1 (425) 722-7703; rozzie@microsoft.com

Mitchell Kapor, President & Chair, Open Source Applications Foundation, 1 (415) 946-3016; fax, 1 (415) 946-3001;
mitch@osafoundation.org

Lisa Dusseault, Lead Designer, Open Source Applications Foundation, lisa@osafoundation.org

Joyce Park, Co-founder & CTO, Renkoo, 1 (650) 315-7241; jp@renkoo.net

Adam Rifkin, Co-founder & CEO, Renkoo, 1 (650) 906-4652; rifkin@renkoo.net

Omar Tawakol, Senior VP, Marketing, Revenue Science, 1 (425) 216-1827; fax, 1 (425) 216-1777;
omart@revenuescience.com

Munjal Shah, Co-founder & CEO, Riya, 1 (650) 278-2304; fax, 1 (650) 887-0375; mshah@riya.com

Steven Hofmeyr, Founder, Sana Security, 1 (650) 292-7100; steve@nthworld.net

Tim Eades, Senior VP, Sales & Marketing, Sana Security, 1 (650) 292-7100; fax, 1 (650) 292-7110;
tim.eades@sanasecurity.com

Ramesh Jain, Chairman, Seraja/EventWeb, 1 (949) 824-0133; fax, 1(949) 824-4056; jain@ics.uci.edu

Tantek Celik, Chief Technologist, Technorati, 1 (650) 335-9558; tantek@technorati.com

Yori Nelken, CEO, TimeBridge, 1 (650) 472-3655; ynelken@timebridge.com

Jeremy Jaech, CEO, Trumba, 1 (206) 625 2262, x106; fax, 1 (206) 625-1437; jeremy@corp.trumba.com

John Arenas, CEO, Worktopia, 1 (914) 468-0811; fax, 1 (914) 931-2180; john.arenas@worktopia.com

Andrew Baio, Co-founder, Upcoming.org, Yahoo!, andy@upcoming.org

Satish Dharmaraj, Co-founder & CEO, Zimbra, 1 (650) 212-0505; fax, 1(650) 212-0504; satish@zimbra.com

Ethan Stock, CEO, Zvents, 1 (650) 234-9629; fax, 1 (650) 234-8322; ethan@zvents.com

Further reading:

Unified Activity Management project: http://www.research.ibm.com/uam

HTML for Calendars, from Foo Camp: http://wiki.oreillynet.com/foocamp04/index.cgi?HTMLForCalendars
hCalendar: http://microformats.org/wiki/hcalendar

IBM's Activity Explorer: http://www-128.ibm.com/developerworks/lotus/library/ae/

DECEMBER 2005 RELEASE 1.0 59

Visit our website: More (free-to-read) columns, ideas, essays and contributors...

featuring Rafe’s Radar, a biweekly column by Rafe Needleman.

http://www.release1-O.com

Release 1.0 Subscription Form

Complete this form and join the other industry executives who regularly rely on Release 1.0 to stay ahead of the headlines. Or if

you wish, you can also subscribe online at www.release1-O.com.

Your annual Release 1.0 subscription costs $285 per year, and includes both the print and electronic versions of 4 quarterly

issues; 25% off the cover price when you order from our online archives; a Release 1.0 binder; a discount on attendance to our

live events; and membership to our website, which includes access to all site content and discussions.

NAME

TITLE COMPANY

ADDRESS

CITY STATE ZIP

COUNTRY

TELEPHONE

FAX

E-MAIL*

URL

*personal e-mail address required for electronic access.

My colleagues should read Release 1.0, too!
Send me information about multiple copy subscriptions and electronic site licenses.

Check enclosed Charge my (circle one): AMERICAN EXPRESS

CARD NUMBER

NAME AND BILLING ADDRESS

SIGNATURE

MASTER CARD VISA

EXPIRATION DATE

Please fax this form to Brodie Crawford at 1 (212) 924-0240.

Payment must be included with this form. Your satisfaction is guaranteed or your money back.

If you wish to pay by check, please mail this form with payment to: Release 1.0, 104 Fifth Avenue, 20th Floor, New York, NY 10011,

USA. If you have any questions, please call us at 1 (212) 924-8800; e-mail us@release1-0.com; www.release1-O.com.

60 RELEASE 1.0

WWW.RELEASE1-0.COM

12-05